Hugging Face's logo

Datasets:
rigbase
/
rigbase-local-llm-benchmarks

Modalities:
Tabular
Text
Formats:
csv
Languages:
English
Size:
< 1K
Tags:
llm
benchmark
gpu
local-llm
llama.cpp
hardware
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
1
rigbase-local-llm-benchmarks / builds.csv
rigbase's picture
rigbase
Initial release: 117 builds, 298 measured results (2026-10-24)
2e7ebaf verified
Raw History Blame Contribute Delete
115 kB
build_slug,build_title,gpus,vram_gb,cpu,ram,power,model,quant,engine,tps,spec_decode,prefill_tps,context,how_measured,status,source_checked_by_rigbase,source,page_url
dual-3090-ryzen-3700x,Dual RTX 3090 on Ryzen 3700X (X470),2x Zotac RTX 3090,48,AMD Ryzen 7 3700X,48 GB DDR4,GPU power limit 270 W each,Llama 3 70B,Q4_K_M,Ollama,16.95,none,,,,,no,https://www.reddit.com/r/LocalLLaMA/comments/1d5vxe7/so_i_bought_second_3090_here_are_my_results_llama/,https://getrigbase.com/builds/dual-3090-ryzen-3700x/
dual-3090-ryzen-3700x,Dual RTX 3090 on Ryzen 3700X (X470),2x Zotac RTX 3090,48,AMD Ryzen 7 3700X,48 GB DDR4,GPU power limit 270 W each,Llama 3 70B,AWQ 4-bit,vLLM (TP=2),21.2,none,,,,,no,https://www.reddit.com/r/LocalLLaMA/comments/1d5vxe7/so_i_bought_second_3090_here_are_my_results_llama/,https://getrigbase.com/builds/dual-3090-ryzen-3700x/
dual-3090-2000-eur-exllamav2,"Dual used RTX 3090, ~2000 EUR total",2x RTX 3090 (used),48,,,GPU power limit 275 W,Llama 3.3 70B,,ExLlamaV2 (TP),20,none,,,,,no,https://www.reddit.com/r/LocalLLaMA/comments/1hdpvpt/llama_33_speed/,https://getrigbase.com/builds/dual-3090-2000-eur-exllamav2/
mac-studio-64gb-llama33,Mac Studio 64 GB,Apple unified memory,64,Apple Silicon (Mac Studio),64 GB unified,,Llama 3.3 70B,Q4,,11,none,,,,,no,https://www.reddit.com/r/LocalLLaMA/comments/1hdpvpt/llama_33_speed/,https://getrigbase.com/builds/mac-studio-64gb-llama33/
3090-plus-4060ti-16gb,RTX 3090 + RTX 4060 Ti 16GB mixed pair,RTX 3090 + RTX 4060 Ti 16GB,40,,,,Llama 3.3 70B,IQ4_XS,,11,none,,,,,no,https://www.reddit.com/r/LocalLLaMA/comments/1hdpvpt/llama_33_speed/,https://getrigbase.com/builds/3090-plus-4060ti-16gb/
single-4060ti-16gb-64gb-ram,Single RTX 4060 Ti 16GB + 64 GB RAM (partial offload),RTX 4060 Ti 16GB,16,,64 GB,,Llama 3.3 70B,Q5_K_M,LM Studio / Kobold,1.5,none,,,,,no,https://www.reddit.com/r/LocalLLaMA/comments/1hdpvpt/llama_33_speed/,https://getrigbase.com/builds/single-4060ti-16gb-64gb-ram/
rtx-4090-llamacpp-cuda,Single RTX 4090 (llama.cpp CUDA scoreboard),RTX 4090 24GB,24,,,,Llama 2 7B,Q4_0,llama.cpp CUDA (FA),188.96,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/15013,https://getrigbase.com/builds/rtx-4090-llamacpp-cuda/
rtx-4090-koboldcpp-comparison,RTX 4090 desktop (speed comparison vs Macs),RTX 4090 24GB,24,,,,Llama 3.1 8B,Q8,llama.cpp-based,52.99,none,,,,,yes,https://www.someoddcodeguy.dev/low-context-speed-comparison-macbook-mac-studios-and-rtx-4090/,https://getrigbase.com/builds/rtx-4090-koboldcpp-comparison/
rtx-4090-koboldcpp-comparison,RTX 4090 desktop (speed comparison vs Macs),RTX 4090 24GB,24,,,,Mistral Small 22B,Q6_K,llama.cpp-based,26.72,none,,,,,yes,https://www.someoddcodeguy.dev/low-context-speed-comparison-macbook-mac-studios-and-rtx-4090/,https://getrigbase.com/builds/rtx-4090-koboldcpp-comparison/
m2-ultra-76c-192gb,"Mac Studio M2 Ultra 76-core GPU, 192 GB",Apple M2 Ultra 76-core GPU,192,Apple M2 Ultra (16+8 CPU),192 GB unified,,Llama 2 7B,Q4_0,"llama.cpp Metal (FA, Aug 2026 build)",125.21,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/4167,https://getrigbase.com/builds/m2-ultra-76c-192gb/
m2-ultra-76c-192gb,"Mac Studio M2 Ultra 76-core GPU, 192 GB",Apple M2 Ultra 76-core GPU,192,Apple M2 Ultra (16+8 CPU),192 GB unified,,gpt-oss 20B,MXFP4,llama.cpp,116.08,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/4167,https://getrigbase.com/builds/m2-ultra-76c-192gb/
m2-ultra-76c-192gb,"Mac Studio M2 Ultra 76-core GPU, 192 GB",Apple M2 Ultra 76-core GPU,192,Apple M2 Ultra (16+8 CPU),192 GB unified,,gpt-oss 120B,MXFP4,llama.cpp,79.68,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/4167,https://getrigbase.com/builds/m2-ultra-76c-192gb/
macbook-pro-m2-max,MacBook Pro M2 Max,Apple M2 Max,,Apple M2 Max,,,Llama 3.1 8B,Q8,llama.cpp-based,28.92,none,,,,,yes,https://www.someoddcodeguy.dev/low-context-speed-comparison-macbook-mac-studios-and-rtx-4090/,https://getrigbase.com/builds/macbook-pro-m2-max/
mac-studio-m1-ultra,Mac Studio M1 Ultra,Apple M1 Ultra,,Apple M1 Ultra,,,Llama 3.1 8B,Q8,llama.cpp-based,37.92,none,,,,,yes,https://www.someoddcodeguy.dev/low-context-speed-comparison-macbook-mac-studios-and-rtx-4090/,https://getrigbase.com/builds/mac-studio-m1-ultra/
macbook-m3-max-40c-64gb,"MacBook Pro M3 Max 40-core GPU, 64 GB",Apple M3 Max 40-core GPU,64,Apple M3 Max,64 GB unified,,Llama 3 8B,Q4_K_M,llama.cpp,50.74,none,,,,,yes,https://github.com/XiongjieDai/GPU-Benchmarks-on-LLM-Inference,https://getrigbase.com/builds/macbook-m3-max-40c-64gb/
macbook-m1-max-32c-64gb,"MacBook Pro M1 Max 32-core GPU, 64 GB",Apple M1 Max 32-core GPU,64,Apple M1 Max,64 GB unified,,Llama 3 8B,Q4_K_M,llama.cpp,34.49,none,,,,,yes,https://github.com/XiongjieDai/GPU-Benchmarks-on-LLM-Inference,https://getrigbase.com/builds/macbook-m1-max-32c-64gb/
epyc-512gb-cpu-only-deepseek,"$2K EPYC 64-core, 512 GB DDR4, CPU-only",None (CPU inference),0,AMD EPYC 64-core (7702 / 7C13 / 7V13),512 GB DDR4,,DeepSeek R1 671B,Q4,llama.cpp (bare metal Ubuntu 24),4.31,none,,,,,yes,https://www.youtube.com/watch?v=Tq_cmN4j2yY,https://getrigbase.com/builds/epyc-512gb-cpu-only-deepseek/
m3-ultra-512gb-deepseek,"Mac Studio M3 Ultra, 512 GB",Apple M3 Ultra,512,Apple M3 Ultra,512 GB unified,,DeepSeek R1 671B,Q4,MLX,18.43,none,,,,,no,https://www.reddit.com/r/LocalLLaMA/comments/1j9vjf1/deepseek_r1_671b_q4_m3_ultra_512gb_with_mlx/,https://getrigbase.com/builds/m3-ultra-512gb-deepseek/
framework-desktop-strix-halo,Framework Desktop (AMD Strix Halo) mainboard,"AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU; Ollama runs CPU only, llama.cpp runs on iGPU via Vulkan)",,AMD Ryzen AI Max (Strix Halo),,"~141 W during Ollama CPU runs; ~133 W during llama.cpp Vulkan iGPU runs (Llama 3.2 3B, Llama 3.1 70B)",DeepSeek R1 Distill 8B,Ollama default,"Ollama (CPU only, iGPU not used)",19.49,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/21,https://getrigbase.com/builds/framework-desktop-strix-halo/
framework-desktop-strix-halo,Framework Desktop (AMD Strix Halo) mainboard,"AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU; Ollama runs CPU only, llama.cpp runs on iGPU via Vulkan)",,AMD Ryzen AI Max (Strix Halo),,"~141 W during Ollama CPU runs; ~133 W during llama.cpp Vulkan iGPU runs (Llama 3.2 3B, Llama 3.1 70B)",DeepSeek R1 Distill 70B,Ollama default,"Ollama (CPU only, iGPU not used)",2.43,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/21,https://getrigbase.com/builds/framework-desktop-strix-halo/
framework-desktop-strix-halo,Framework Desktop (AMD Strix Halo) mainboard,"AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU; Ollama runs CPU only, llama.cpp runs on iGPU via Vulkan)",,AMD Ryzen AI Max (Strix Halo),,"~141 W during Ollama CPU runs; ~133 W during llama.cpp Vulkan iGPU runs (Llama 3.2 3B, Llama 3.1 70B)",Llama 3.1 70B,Q4_K_M,"llama.cpp Vulkan (iGPU, llama-bench tg128)",4.97,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/21,https://getrigbase.com/builds/framework-desktop-strix-halo/
framework-desktop-strix-halo,Framework Desktop (AMD Strix Halo) mainboard,"AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU; Ollama runs CPU only, llama.cpp runs on iGPU via Vulkan)",,AMD Ryzen AI Max (Strix Halo),,"~141 W during Ollama CPU runs; ~133 W during llama.cpp Vulkan iGPU runs (Llama 3.2 3B, Llama 3.1 70B)",Llama 3.2 3B,Q4_K_M,"llama.cpp Vulkan (iGPU, llama-bench tg128)",88.14,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/21,https://getrigbase.com/builds/framework-desktop-strix-halo/
4x3090-tr32-192gb-qwen3-235b,"4x RTX 3090 + 32-core Threadripper, 192 GB DDR4",4x RTX 3090,96,AMD Threadripper 32-core,192 GB DDR4,,Qwen3 235B-A22B,UD-Q3_K_XL,llama.cpp,12-14,none,,,,,no,https://www.reddit.com/r/LocalLLaMA/comments/1khmaah/5_commands_to_run_qwen3235ba22b_q3_inference_on/,https://getrigbase.com/builds/4x3090-tr32-192gb-qwen3-235b/
4x3090-epyc-7443-256gb,"4x RTX 3090 + EPYC 7443, 256 GB DDR4 8-channel",4x RTX 3090,96,AMD EPYC 7443 (24 cores),256 GB DDR4 3200 (8 channel),,Qwen3 235B-A22B,IQ4_XS,llama.cpp,18,none,,,,,no,https://www.reddit.com/r/LocalLLaMA/comments/1khmaah/5_commands_to_run_qwen3235ba22b_q3_inference_on/,https://getrigbase.com/builds/4x3090-epyc-7443-256gb/
epyc-7f52-6gpu-136gb,EPYC 7F52 jank rig: 4x3090 + 4090 + 4080 (136 GB VRAM),"4x RTX 3090, RTX 4090, RTX 4080",136,AMD EPYC 7F52,32 GB RDIMM 3200,,Llama 3.1 70B,Q8_0,Oobabooga,9,none,,,,,no,https://www.reddit.com/r/LocalLLaMA/comments/1fqwler/show_me_your_ai_rig/,https://getrigbase.com/builds/epyc-7f52-6gpu-136gb/
epyc-plus-3090-deepseek-q2,EPYC + single RTX 3090 (hybrid),RTX 3090,24,AMD EPYC,,,DeepSeek R1 671B,Q2_K,llama.cpp CUDA,8.97,none,,,,,no,https://www.reddit.com/r/LocalLLaMA/comments/1j2c7b4/unsloth_ds_r1_222b_2_ts_on_dell_r730_cpu_help/,https://getrigbase.com/builds/epyc-plus-3090-deepseek-q2/
intel-265k-rtx-4090,Intel Core Ultra 265K + RTX 4090,RTX 4090,24,Intel Core Ultra 7 265K,,"Peak 509.1 W (gpt-oss 20B), 464.2 W (DeepSeek R1 14B)",gpt-oss 20B,,llama.cpp,279.65,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/43,https://getrigbase.com/builds/intel-265k-rtx-4090/
intel-265k-rtx-4090,Intel Core Ultra 265K + RTX 4090,RTX 4090,24,Intel Core Ultra 7 265K,,"Peak 509.1 W (gpt-oss 20B), 464.2 W (DeepSeek R1 14B)",DeepSeek R1 Distill 14B,,llama.cpp,92.49,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/43,https://getrigbase.com/builds/intel-265k-rtx-4090/
intel-265k-rtx-4090,Intel Core Ultra 265K + RTX 4090,RTX 4090,24,Intel Core Ultra 7 265K,,"Peak 509.1 W (gpt-oss 20B), 464.2 W (DeepSeek R1 14B)",Llama 3.2 3B,,llama.cpp,334.67,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/43,https://getrigbase.com/builds/intel-265k-rtx-4090/
intel-265k-rtx-3080-ti,Intel Core Ultra 265K + RTX 3080 Ti,RTX 3080 Ti,12,Intel Core Ultra 7 265K,,Peak 495 W,DeepSeek R1 Distill 14B,,llama.cpp,70.74,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/39,https://getrigbase.com/builds/intel-265k-rtx-3080-ti/
intel-265k-rtx-3080-ti,Intel Core Ultra 265K + RTX 3080 Ti,RTX 3080 Ti,12,Intel Core Ultra 7 265K,,Peak 495 W,Llama 3.2 3B,,llama.cpp,263.24,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/39,https://getrigbase.com/builds/intel-265k-rtx-3080-ti/
intel-265k-radeon-ai-pro-r9700,Intel Core Ultra 265K + Radeon AI Pro R9700,AMD Radeon AI Pro R9700,32,Intel Core Ultra 7 265K,,Peak 332.9 W (gpt-oss 20B),gpt-oss 20B,,llama.cpp,163.26,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/35,https://getrigbase.com/builds/intel-265k-radeon-ai-pro-r9700/
intel-265k-radeon-ai-pro-r9700,Intel Core Ultra 265K + Radeon AI Pro R9700,AMD Radeon AI Pro R9700,32,Intel Core Ultra 7 265K,,Peak 332.9 W (gpt-oss 20B),DeepSeek R1 Distill 14B,,llama.cpp,53.72,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/35,https://getrigbase.com/builds/intel-265k-radeon-ai-pro-r9700/
intel-265k-radeon-ai-pro-r9700,Intel Core Ultra 265K + Radeon AI Pro R9700,AMD Radeon AI Pro R9700,32,Intel Core Ultra 7 265K,,Peak 332.9 W (gpt-oss 20B),Llama 3.2 3B,,llama.cpp,195.91,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/35,https://getrigbase.com/builds/intel-265k-radeon-ai-pro-r9700/
intel-265k-rtx-4070-ti,Intel Core Ultra 265K + RTX 4070 Ti,RTX 4070 Ti,12,Intel Core Ultra 7 265K,,Peak 298.7 W (gpt-oss 20B),gpt-oss 20B,,llama.cpp,163.14,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/45,https://getrigbase.com/builds/intel-265k-rtx-4070-ti/
intel-265k-rtx-4070-ti,Intel Core Ultra 265K + RTX 4070 Ti,RTX 4070 Ti,12,Intel Core Ultra 7 265K,,Peak 298.7 W (gpt-oss 20B),DeepSeek R1 Distill 14B,,llama.cpp,49.74,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/45,https://getrigbase.com/builds/intel-265k-rtx-4070-ti/
intel-265k-rtx-4070-ti,Intel Core Ultra 265K + RTX 4070 Ti,RTX 4070 Ti,12,Intel Core Ultra 7 265K,,Peak 298.7 W (gpt-oss 20B),Llama 3.2 3B,,llama.cpp,196.15,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/45,https://getrigbase.com/builds/intel-265k-rtx-4070-ti/
intel-265k-rtx-2080-ti,Intel Core Ultra 265K + RTX 2080 Ti,RTX 2080 Ti,11,Intel Core Ultra 7 265K,,Peak 307.8 W (DeepSeek R1 14B),DeepSeek R1 Distill 14B,,llama.cpp,45.06,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/46,https://getrigbase.com/builds/intel-265k-rtx-2080-ti/
intel-265k-rtx-2080-ti,Intel Core Ultra 265K + RTX 2080 Ti,RTX 2080 Ti,11,Intel Core Ultra 7 265K,,Peak 307.8 W (DeepSeek R1 14B),Llama 3.2 3B,,llama.cpp,180.38,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/46,https://getrigbase.com/builds/intel-265k-rtx-2080-ti/
intel-265k-rtx-3060,Intel Core Ultra 265K + RTX 3060,RTX 3060,12,Intel Core Ultra 7 265K,,Peak 229.3 W (gpt-oss 20B),gpt-oss 20B,,llama.cpp,83.96,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/40,https://getrigbase.com/builds/intel-265k-rtx-3060/
intel-265k-rtx-3060,Intel Core Ultra 265K + RTX 3060,RTX 3060,12,Intel Core Ultra 7 265K,,Peak 229.3 W (gpt-oss 20B),DeepSeek R1 Distill 14B,,llama.cpp,29.77,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/40,https://getrigbase.com/builds/intel-265k-rtx-3060/
intel-265k-rtx-3060,Intel Core Ultra 265K + RTX 3060,RTX 3060,12,Intel Core Ultra 7 265K,,Peak 229.3 W (gpt-oss 20B),Llama 3.2 3B,,llama.cpp,122.85,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/40,https://getrigbase.com/builds/intel-265k-rtx-3060/
intel-265k-rtx-a4000,Intel Core Ultra 265K + RTX A4000,RTX A4000,16,Intel Core Ultra 7 265K,,Peak 199.4 W (gpt-oss 20B),gpt-oss 20B,,llama.cpp,93.55,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/36,https://getrigbase.com/builds/intel-265k-rtx-a4000/
intel-265k-rtx-a4000,Intel Core Ultra 265K + RTX A4000,RTX A4000,16,Intel Core Ultra 7 265K,,Peak 199.4 W (gpt-oss 20B),Llama 3.2 3B,,llama.cpp,149.44,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/36,https://getrigbase.com/builds/intel-265k-rtx-a4000/
mac-studio-m3-ultra-512gb-geerling,Mac Studio M3 Ultra 512GB (Geerling test),Apple M3 Ultra (unified),512,Apple M3 Ultra,,"Peak 261.8 W (DeepSeek R1 671B), 243 W (Llama 3.1 70B), 227 W (gpt-oss 20B), 223 W (Llama 3.2 3B)",DeepSeek R1 671B,,llama.cpp,19.89,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/42,https://getrigbase.com/builds/mac-studio-m3-ultra-512gb-geerling/
mac-studio-m3-ultra-512gb-geerling,Mac Studio M3 Ultra 512GB (Geerling test),Apple M3 Ultra (unified),512,Apple M3 Ultra,,"Peak 261.8 W (DeepSeek R1 671B), 243 W (Llama 3.1 70B), 227 W (gpt-oss 20B), 223 W (Llama 3.2 3B)",gpt-oss 20B,,llama.cpp,115.29,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/42,https://getrigbase.com/builds/mac-studio-m3-ultra-512gb-geerling/
mac-studio-m3-ultra-512gb-geerling,Mac Studio M3 Ultra 512GB (Geerling test),Apple M3 Ultra (unified),512,Apple M3 Ultra,,"Peak 261.8 W (DeepSeek R1 671B), 243 W (Llama 3.1 70B), 227 W (gpt-oss 20B), 223 W (Llama 3.2 3B)",gpt-oss 20B,MXFP4,"llama.cpp (Metal, llama-bench)",115.52,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/42,https://getrigbase.com/builds/mac-studio-m3-ultra-512gb-geerling/
mac-studio-m3-ultra-512gb-geerling,Mac Studio M3 Ultra 512GB (Geerling test),Apple M3 Ultra (unified),512,Apple M3 Ultra,,"Peak 261.8 W (DeepSeek R1 671B), 243 W (Llama 3.1 70B), 227 W (gpt-oss 20B), 223 W (Llama 3.2 3B)",Llama 3.1 70B,,llama.cpp,14.08,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/42,https://getrigbase.com/builds/mac-studio-m3-ultra-512gb-geerling/
mac-studio-m3-ultra-512gb-geerling,Mac Studio M3 Ultra 512GB (Geerling test),Apple M3 Ultra (unified),512,Apple M3 Ultra,,"Peak 261.8 W (DeepSeek R1 671B), 243 W (Llama 3.1 70B), 227 W (gpt-oss 20B), 223 W (Llama 3.2 3B)",Llama 3.1 405B,,llama.cpp,1.96,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/42,https://getrigbase.com/builds/mac-studio-m3-ultra-512gb-geerling/
mac-studio-m3-ultra-512gb-geerling,Mac Studio M3 Ultra 512GB (Geerling test),Apple M3 Ultra (unified),512,Apple M3 Ultra,,"Peak 261.8 W (DeepSeek R1 671B), 243 W (Llama 3.1 70B), 227 W (gpt-oss 20B), 223 W (Llama 3.2 3B)",DeepSeek R1 Distill 14B,,llama.cpp,51.85,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/42,https://getrigbase.com/builds/mac-studio-m3-ultra-512gb-geerling/
mac-studio-m3-ultra-512gb-geerling,Mac Studio M3 Ultra 512GB (Geerling test),Apple M3 Ultra (unified),512,Apple M3 Ultra,,"Peak 261.8 W (DeepSeek R1 671B), 243 W (Llama 3.1 70B), 227 W (gpt-oss 20B), 223 W (Llama 3.2 3B)",Llama 3.2 3B,,llama.cpp,154.6,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/42,https://getrigbase.com/builds/mac-studio-m3-ultra-512gb-geerling/
mac-studio-m1-ultra-64c-128gb,Mac Studio M1 Ultra (64-core GPU) 128GB,Apple M1 Ultra 64-core GPU (unified),128,Apple M1 Ultra,,,gpt-oss 20B,,llama.cpp,100.43,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/48,https://getrigbase.com/builds/mac-studio-m1-ultra-64c-128gb/
mac-studio-m1-ultra-64c-128gb,Mac Studio M1 Ultra (64-core GPU) 128GB,Apple M1 Ultra 64-core GPU (unified),128,Apple M1 Ultra,,,DeepSeek R1 Distill 14B,,llama.cpp,41.51,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/48,https://getrigbase.com/builds/mac-studio-m1-ultra-64c-128gb/
mac-studio-m1-ultra-64c-128gb,Mac Studio M1 Ultra (64-core GPU) 128GB,Apple M1 Ultra 64-core GPU (unified),128,Apple M1 Ultra,,,Llama 3.1 70B,,llama.cpp,9.84,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/48,https://getrigbase.com/builds/mac-studio-m1-ultra-64c-128gb/
mac-studio-m1-ultra-64c-128gb,Mac Studio M1 Ultra (64-core GPU) 128GB,Apple M1 Ultra 64-core GPU (unified),128,Apple M1 Ultra,,,Llama 3.2 3B,,llama.cpp,127.6,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/48,https://getrigbase.com/builds/mac-studio-m1-ultra-64c-128gb/
dell-pro-max-gb10-dgx-spark,Dell Pro Max with GB10 (Nvidia DGX Spark),Nvidia GB10 (unified),,Nvidia GB10,,"Peak 122.8 W (gpt-oss 20B), 128.5 W (Llama 3.2 3B), 156 W (Llama 3.1 70B)",gpt-oss 20B,,llama.cpp,87.97,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/34,https://getrigbase.com/builds/dell-pro-max-gb10-dgx-spark/
dell-pro-max-gb10-dgx-spark,Dell Pro Max with GB10 (Nvidia DGX Spark),Nvidia GB10 (unified),,Nvidia GB10,,"Peak 122.8 W (gpt-oss 20B), 128.5 W (Llama 3.2 3B), 156 W (Llama 3.1 70B)",DeepSeek R1 Distill 14B,,llama.cpp,23.92,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/34,https://getrigbase.com/builds/dell-pro-max-gb10-dgx-spark/
dell-pro-max-gb10-dgx-spark,Dell Pro Max with GB10 (Nvidia DGX Spark),Nvidia GB10 (unified),,Nvidia GB10,,"Peak 122.8 W (gpt-oss 20B), 128.5 W (Llama 3.2 3B), 156 W (Llama 3.1 70B)",Llama 3.1 70B,,llama.cpp,4.71,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/34,https://getrigbase.com/builds/dell-pro-max-gb10-dgx-spark/
dell-pro-max-gb10-dgx-spark,Dell Pro Max with GB10 (Nvidia DGX Spark),Nvidia GB10 (unified),,Nvidia GB10,,"Peak 122.8 W (gpt-oss 20B), 128.5 W (Llama 3.2 3B), 156 W (Llama 3.1 70B)",Llama 3.2 3B,,llama.cpp,97.93,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/34,https://getrigbase.com/builds/dell-pro-max-gb10-dgx-spark/
intel-13900k-rtx-5090,Intel i9-13900K + RTX 5090,RTX 5090,32,Intel Core i9-13900K,,,Llama 3.2 3B,,llama.cpp,271.4,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/pull/18,https://getrigbase.com/builds/intel-13900k-rtx-5090/
ryzen-7900x-rtx-4090,Ryzen 9 7900X + RTX 4090,RTX 4090,24,AMD Ryzen 9 7900X,,,Llama 3.2 3B,,llama.cpp,237.05,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/pull/11,https://getrigbase.com/builds/ryzen-7900x-rtx-4090/
ryzen-7900x-rtx-4090,Ryzen 9 7900X + RTX 4090,RTX 4090,24,AMD Ryzen 9 7900X,,,Llama 3.1 70B,,GPU+CPU offload,3.1,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/pull/11,https://getrigbase.com/builds/ryzen-7900x-rtx-4090/
ryzen-7950x-rtx-4080,Ryzen 9 7950X + RTX 4080,RTX 4080,16,AMD Ryzen 9 7950X,,,Llama 3.2 3B,,llama.cpp,204.45,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/pull/11,https://getrigbase.com/builds/ryzen-7950x-rtx-4080/
ryzen-7950x-rtx-4070-ti-super,Ryzen 9 7950X + RTX 4070 Ti Super,RTX 4070 Ti Super,16,AMD Ryzen 9 7950X,,,Llama 3.2 3B,,llama.cpp,198.95,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/pull/11,https://getrigbase.com/builds/ryzen-7950x-rtx-4070-ti-super/
ryzen-9950x-rx-7900-xt,Ryzen 9 9950X + Radeon RX 7900 XT,AMD Radeon RX 7900 XT,20,AMD Ryzen 9 9950X,,,Llama 3.2 3B,,llama.cpp,131.2,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/pull/11,https://getrigbase.com/builds/ryzen-9950x-rx-7900-xt/
m4-mac-mini-32gb,M4 Mac mini (10-core) 32GB,Apple M4 (unified),32,Apple M4,,Peak 30.1 W,Llama 3.2 3B,,llama.cpp,41.31,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/2,https://getrigbase.com/builds/m4-mac-mini-32gb/
ampereone-a192-512gb-cpu,AmpereOne A192-32X 512GB (CPU only),None (CPU only),,AmpereOne A192-32X,512 GB,Peak 477 W (DeepSeek R1 671B),DeepSeek R1 671B,,llama.cpp,4.18,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/10,https://getrigbase.com/builds/ampereone-a192-512gb-cpu/
ampereone-a192-512gb-cpu,AmpereOne A192-32X 512GB (CPU only),None (CPU only),,AmpereOne A192-32X,512 GB,Peak 477 W (DeepSeek R1 671B),Llama 3.1 70B,,llama.cpp,3.86,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/10,https://getrigbase.com/builds/ampereone-a192-512gb-cpu/
ampereone-a192-512gb-cpu,AmpereOne A192-32X 512GB (CPU only),None (CPU only),,AmpereOne A192-32X,512 GB,Peak 477 W (DeepSeek R1 671B),Llama 3.1 405B,,llama.cpp,0.9,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/10,https://getrigbase.com/builds/ampereone-a192-512gb-cpu/
ampereone-a192-512gb-cpu,AmpereOne A192-32X 512GB (CPU only),None (CPU only),,AmpereOne A192-32X,512 GB,Peak 477 W (DeepSeek R1 671B),Llama 3.2 3B,,llama.cpp,23.52,none,,,,,yes,https://github.com/geerlingguy/ai-benchmarks/issues/10,https://getrigbase.com/builds/ampereone-a192-512gb-cpu/
macbook-m4-max-36gb,Apple M4 Max 36GB,Apple M4 Max (unified),36,Apple M4 Max,,,gpt-oss 20B,MXFP4,llama.cpp,92.36,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/15396,https://getrigbase.com/builds/macbook-m4-max-36gb/
apple-m1-max-64gb-gpt-oss,Apple M1 Max 64GB,Apple M1 Max (unified),64,Apple M1 Max,,,gpt-oss 20B,MXFP4,llama.cpp,75.15,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/15396,https://getrigbase.com/builds/apple-m1-max-64gb-gpt-oss/
rtx-pro-6000-96gb,RTX Pro 6000 96GB workstation card,RTX Pro 6000 (Blackwell),96,,,,gpt-oss 120B,MXFP4,llama.cpp,196.31,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/15396,https://getrigbase.com/builds/rtx-pro-6000-96gb/
rtx-pro-6000-96gb,RTX Pro 6000 96GB workstation card,RTX Pro 6000 (Blackwell),96,,,,gpt-oss 20B,MXFP4,llama.cpp,286.91,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/15396,https://getrigbase.com/builds/rtx-pro-6000-96gb/
single-rtx-3090-gpt-oss,Single RTX 3090 24GB,RTX 3090,24,,,,gpt-oss 20B,MXFP4,llama.cpp,161.77,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/15396,https://getrigbase.com/builds/single-rtx-3090-gpt-oss/
single-rtx-5090-gpt-oss,Single RTX 5090 32GB,RTX 5090,32,,,,gpt-oss 20B,MXFP4,llama.cpp,282.51,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/15396,https://getrigbase.com/builds/single-rtx-5090-gpt-oss/
single-rtx-5070-ti-gpt-oss,Single RTX 5070 Ti 16GB,RTX 5070 Ti,16,,,,gpt-oss 20B,MXFP4,llama.cpp,189.45,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/15396,https://getrigbase.com/builds/single-rtx-5070-ti-gpt-oss/
single-rtx-5060-ti-16gb-gpt-oss,Single RTX 5060 Ti 16GB,RTX 5060 Ti 16GB,16,,,,gpt-oss 20B,MXFP4,llama.cpp,111.51,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/15396,https://getrigbase.com/builds/single-rtx-5060-ti-16gb-gpt-oss/
single-rtx-5080-gpt-oss,Single RTX 5080 16GB,RTX 5080,16,,,,gpt-oss 20B,MXFP4,llama.cpp,204.85,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/15396,https://getrigbase.com/builds/single-rtx-5080-gpt-oss/
llamabench-rtx-5090,RTX 5090: 2026 models (llamabench community),RTX 5090,32,,,,Qwen3.6 35B-A3B,llama-bench tg128 (quant per result),llama.cpp,258,none,,,,,yes,https://llamabench.ai/browse/rtx5090,https://getrigbase.com/builds/llamabench-rtx-5090/
llamabench-rtx-5090,RTX 5090: 2026 models (llamabench community),RTX 5090,32,,,,Qwen3.6 27B,llama-bench tg128 (quant per result),llama.cpp,75.7,none,,,,,yes,https://llamabench.ai/browse/rtx5090,https://getrigbase.com/builds/llamabench-rtx-5090/
llamabench-rtx-5090,RTX 5090: 2026 models (llamabench community),RTX 5090,32,,,,Gemma 4 31B,llama-bench tg128 (quant per result),llama.cpp,70.0,none,,,,,yes,https://llamabench.ai/browse/rtx5090,https://getrigbase.com/builds/llamabench-rtx-5090/
llamabench-rtx-5090,RTX 5090: 2026 models (llamabench community),RTX 5090,32,,,,Gemma 4 26B-A4B,llama-bench tg128 (quant per result),llama.cpp,235,none,,,,,yes,https://llamabench.ai/browse/rtx5090,https://getrigbase.com/builds/llamabench-rtx-5090/
llamabench-rtx-5090,RTX 5090: 2026 models (llamabench community),RTX 5090,32,,,,Gemma 4 12B,llama-bench tg128 (quant per result),llama.cpp,142,none,,,,,yes,https://llamabench.ai/browse/rtx5090,https://getrigbase.com/builds/llamabench-rtx-5090/
llamabench-rtx-5090,RTX 5090: 2026 models (llamabench community),RTX 5090,32,,,,gpt-oss 20B,llama-bench tg128 (quant per result),llama.cpp,313,none,,,,,yes,https://llamabench.ai/browse/rtx5090,https://getrigbase.com/builds/llamabench-rtx-5090/
llamabench-rtx-5090,RTX 5090: 2026 models (llamabench community),RTX 5090,32,,,,Qwen3.5 9B,llama-bench tg128 (quant per result),llama.cpp,208,none,,,,,yes,https://llamabench.ai/browse/rtx5090,https://getrigbase.com/builds/llamabench-rtx-5090/
llamabench-rtx-4090,RTX 4090: 2026 models (llamabench community),RTX 4090,24,,,,Qwen3.6 27B,llama-bench tg128 (quant per result),llama.cpp,47.0,none,,,,,yes,https://llamabench.ai/browse/rtx4090,https://getrigbase.com/builds/llamabench-rtx-4090/
llamabench-rtx-4090,RTX 4090: 2026 models (llamabench community),RTX 4090,24,,,,Gemma 4 31B,llama-bench tg128 (quant per result),llama.cpp,43.8,none,,,,,yes,https://llamabench.ai/browse/rtx4090,https://getrigbase.com/builds/llamabench-rtx-4090/
llamabench-rtx-4090,RTX 4090: 2026 models (llamabench community),RTX 4090,24,,,,Gemma 4 26B-A4B,llama-bench tg128 (quant per result),llama.cpp,164,none,,,,,yes,https://llamabench.ai/browse/rtx4090,https://getrigbase.com/builds/llamabench-rtx-4090/
llamabench-rtx-4090,RTX 4090: 2026 models (llamabench community),RTX 4090,24,,,,Gemma 4 12B,llama-bench tg128 (quant per result),llama.cpp,98.1,none,,,,,yes,https://llamabench.ai/browse/rtx4090,https://getrigbase.com/builds/llamabench-rtx-4090/
llamabench-rtx-4090,RTX 4090: 2026 models (llamabench community),RTX 4090,24,,,,gpt-oss 20B,llama-bench tg128 (quant per result),llama.cpp,203,none,,,,,yes,https://llamabench.ai/browse/rtx4090,https://getrigbase.com/builds/llamabench-rtx-4090/
llamabench-rtx-4090,RTX 4090: 2026 models (llamabench community),RTX 4090,24,,,,Qwen3.5 9B,llama-bench tg128 (quant per result),llama.cpp,137,none,,,,,yes,https://llamabench.ai/browse/rtx4090,https://getrigbase.com/builds/llamabench-rtx-4090/
llamabench-rtx-4090,RTX 4090: 2026 models (llamabench community),RTX 4090,24,,,,LFM2.5 24B-A2B,llama-bench tg128 (quant per result),llama.cpp,261-339,none,,,,,yes,https://llamabench.ai/browse/rtx4090,https://getrigbase.com/builds/llamabench-rtx-4090/
llamabench-rtx-4090,RTX 4090: 2026 models (llamabench community),RTX 4090,24,,,,Nemotron 3.5 Bolt 74B-A9B,llama-bench tg128 (quant per result),llama.cpp,108,none,,,,,yes,https://llamabench.ai/browse/rtx4090,https://getrigbase.com/builds/llamabench-rtx-4090/
llamabench-rtx-3090,RTX 3090: 2026 models (llamabench community),RTX 3090,24,,,,Qwen3.6 35B-A3B,llama-bench tg128 (quant per result),llama.cpp,146-153,none,,,,,yes,https://llamabench.ai/browse/rtx3090,https://getrigbase.com/builds/llamabench-rtx-3090/
llamabench-rtx-3090,RTX 3090: 2026 models (llamabench community),RTX 3090,24,,,,Qwen3.6 27B,llama-bench tg128 (quant per result),llama.cpp,40.2,none,,,,,yes,https://llamabench.ai/browse/rtx3090,https://getrigbase.com/builds/llamabench-rtx-3090/
llamabench-rtx-3090,RTX 3090: 2026 models (llamabench community),RTX 3090,24,,,,Qwen3.8 27B,llama-bench tg128 (quant per result),llama.cpp,38.2-71.6,none,,,,,yes,https://llamabench.ai/browse/rtx3090,https://getrigbase.com/builds/llamabench-rtx-3090/
llamabench-rtx-5070-ti,RTX 5070 Ti: 2026 models (llamabench community),RTX 5070 Ti,16,,,,Gemma 4 12B,llama-bench tg128 (quant per result),llama.cpp,87.6,none,,,,,yes,https://llamabench.ai/browse/rtx5070ti,https://getrigbase.com/builds/llamabench-rtx-5070-ti/
llamabench-rtx-3060,RTX 3060 12GB: 2026 models (llamabench community),RTX 3060,12,,,,Gemma 4 12B,llama-bench tg128 (quant per result),llama.cpp,38.5,none,,,,,yes,https://llamabench.ai/browse/rtx3060,https://getrigbase.com/builds/llamabench-rtx-3060/
llamabench-rtx-3060,RTX 3060 12GB: 2026 models (llamabench community),RTX 3060,12,,,,LFM2.5 24B-A2B,llama-bench tg128 (quant per result),llama.cpp,125,none,,,,,yes,https://llamabench.ai/browse/rtx3060,https://getrigbase.com/builds/llamabench-rtx-3060/
llamabench-rtx-3060,RTX 3060 12GB: 2026 models (llamabench community),RTX 3060,12,,,,Qwen3.8 27B,llama-bench tg128 (quant per result),llama.cpp,5.2-6.7,none,,,,,yes,https://llamabench.ai/browse/rtx3060,https://getrigbase.com/builds/llamabench-rtx-3060/
llamabench-rtx-4060-ti-16gb,RTX 4060 Ti 16GB: 2026 models (llamabench community),RTX 4060 Ti 16GB,16,,,,LFM2.5 24B-A2B,llama-bench tg128 (quant per result),llama.cpp,131-150,none,,,,,yes,https://llamabench.ai/browse/rtx4060ti16gb,https://getrigbase.com/builds/llamabench-rtx-4060-ti-16gb/
llamabench-rtx-4060-ti-16gb,RTX 4060 Ti 16GB: 2026 models (llamabench community),RTX 4060 Ti 16GB,16,,,,Qwen3.8 27B,llama-bench tg128 (quant per result),llama.cpp,10.7-17.6,none,,,,,yes,https://llamabench.ai/browse/rtx4060ti16gb,https://getrigbase.com/builds/llamabench-rtx-4060-ti-16gb/
llamabench-radeon-8060s-strix-halo,"Strix Halo (Radeon 8060S, 96 GB): 2026 models (llamabench community)",AMD Ryzen AI Max+ (Radeon 8060S iGPU),96,,,,Qwen3.6 35B-A3B,llama-bench tg128 (quant per result),llama.cpp,50.2-62.4,none,,,,,yes,https://llamabench.ai/browse/radeon8060s,https://getrigbase.com/builds/llamabench-radeon-8060s-strix-halo/
rtx-3090-qwen36-35b-a3b-giles,Single RTX 3090: Qwen3.6 35B-A3B (Giles Thomas),RTX 3090,24,,,,Qwen3.6 35B-A3B,"Q4 GGUF, all layers on GPU, ~90K ctx",llama.cpp CUDA,139.6,none,,,,,yes,https://www.gilesthomas.com/2026/07/benchmarking-qwen-3-6-35b-moe-rtx-3090,https://getrigbase.com/builds/rtx-3090-qwen36-35b-a3b-giles/
4x3090-qwen36-vllm-tp2,4x RTX 3090 lab: Qwen3.6 on vLLM vs llama.cpp,4x RTX 3090,96,,,,Qwen3.6 35B-A3B,IQ4_XS (1 GPU),llama.cpp,115-133,none,,,,,yes,https://github.com/tfriedel/qwen3.6-rtx3090-lab,https://getrigbase.com/builds/4x3090-qwen36-vllm-tp2/
4x3090-qwen36-vllm-tp2,4x RTX 3090 lab: Qwen3.6 on vLLM vs llama.cpp,4x RTX 3090,96,,,,Qwen3.6 35B-A3B,AWQ-INT4 (TP=2),vLLM,149,none,,,,,yes,https://github.com/tfriedel/qwen3.6-rtx3090-lab,https://getrigbase.com/builds/4x3090-qwen36-vllm-tp2/
4x3090-qwen36-vllm-tp2,4x RTX 3090 lab: Qwen3.6 on vLLM vs llama.cpp,4x RTX 3090,96,,,,Qwen3.6 35B-A3B,AWQ-INT4 + MTP-3 (TP=2),vLLM,225-231,MTP,,,,,yes,https://github.com/tfriedel/qwen3.6-rtx3090-lab,https://getrigbase.com/builds/4x3090-qwen36-vllm-tp2/
mac-studio-m3-ultra-512gb-muse-spark-open,Mac Studio M3 Ultra 512GB: Muse Spark Open 816B (Hugging Face test),Apple M3 Ultra (unified),512,Apple M3 Ultra,,,Muse Spark Open 816B-A42B,"Q2_K (imatrix), ~304 GB",llama.cpp,18,none,,,,,yes,https://huggingface.co/blog/meta-muse-spark-open,https://getrigbase.com/builds/mac-studio-m3-ultra-512gb-muse-spark-open/
mac-studio-m3-ultra-512gb-muse-spark-open,Mac Studio M3 Ultra 512GB: Muse Spark Open 816B (Hugging Face test),Apple M3 Ultra (unified),512,Apple M3 Ultra,,,Muse Spark Open 816B-A42B,"oQ3e (3-bit), ~344 GB",MLX (oMLX / mlx-lm),21,none,,,,,yes,https://huggingface.co/blog/meta-muse-spark-open,https://getrigbase.com/builds/mac-studio-m3-ultra-512gb-muse-spark-open/
mac-studio-m5-ultra-512gb-apple-mlx,Mac Studio M5 Ultra 512GB: Qwen3.8 27B and Mistral Medium 3.5 (Apple MLX benchmark),Apple M5 Ultra (unified),512,Apple M5 Ultra,,"Peak ~300 W (Apple, 65K-token context)",Qwen3.8 27B,"BF16, 8,192-token context",MLX (mlx_lm.benchmark),19.69,none,,,,,yes,https://machinelearning.apple.com/research/mlx-distributed-inference,https://getrigbase.com/builds/mac-studio-m5-ultra-512gb-apple-mlx/
mac-studio-m5-ultra-512gb-apple-mlx,Mac Studio M5 Ultra 512GB: Qwen3.8 27B and Mistral Medium 3.5 (Apple MLX benchmark),Apple M5 Ultra (unified),512,Apple M5 Ultra,,"Peak ~300 W (Apple, 65K-token context)",Mistral Medium 3.5 128B,"FP8, 8,192-token context",MLX (mlx_lm.benchmark),7.7,none,,,,,yes,https://machinelearning.apple.com/research/mlx-distributed-inference,https://getrigbase.com/builds/mac-studio-m5-ultra-512gb-apple-mlx/
mac-studio-m3-ultra-512gb-apple-mlx,Mac Studio M3 Ultra 512GB: Qwen3.8 27B and Mistral Medium 3.5 (Apple MLX benchmark),Apple M3 Ultra (unified),512,Apple M3 Ultra,,"Peak ~200 W (Apple, 65K-token context)",Qwen3.8 27B,"BF16, 8,192-token context",MLX (mlx_lm.benchmark),12.46,none,,,,,yes,https://machinelearning.apple.com/research/mlx-distributed-inference,https://getrigbase.com/builds/mac-studio-m3-ultra-512gb-apple-mlx/
mac-studio-m3-ultra-512gb-apple-mlx,Mac Studio M3 Ultra 512GB: Qwen3.8 27B and Mistral Medium 3.5 (Apple MLX benchmark),Apple M3 Ultra (unified),512,Apple M3 Ultra,,"Peak ~200 W (Apple, 65K-token context)",Mistral Medium 3.5 128B,"FP8, 8,192-token context",MLX (mlx_lm.benchmark),4.95,none,,,,,yes,https://machinelearning.apple.com/research/mlx-distributed-inference,https://getrigbase.com/builds/mac-studio-m3-ultra-512gb-apple-mlx/
2x-mac-studio-m5-ultra-rdma-cluster,2x Mac Studio M5 Ultra cluster (Thunderbolt 5 RDMA): Muse Spark Open 816B,"2x Apple M5 Ultra (unified, Mac Studio)",,2x Apple M5 Ultra,,,Muse Spark Open 816B-A42B,"NVFP4, 8,192-token context","MLX, 2-node tensor parallel",24.99,none,,,,,yes,https://machinelearning.apple.com/research/mlx-distributed-inference,https://getrigbase.com/builds/2x-mac-studio-m5-ultra-rdma-cluster/
2x-mac-studio-m5-ultra-rdma-cluster,2x Mac Studio M5 Ultra cluster (Thunderbolt 5 RDMA): Muse Spark Open 816B,"2x Apple M5 Ultra (unified, Mac Studio)",,2x Apple M5 Ultra,,,Muse Spark Open 816B-A42B,"NVFP4, 8,192-token context","MLX, 2-node pipeline parallel",15.74,none,,,,,yes,https://machinelearning.apple.com/research/mlx-distributed-inference,https://getrigbase.com/builds/2x-mac-studio-m5-ultra-rdma-cluster/
2x-mac-studio-m5-ultra-rdma-cluster,2x Mac Studio M5 Ultra cluster (Thunderbolt 5 RDMA): Muse Spark Open 816B,"2x Apple M5 Ultra (unified, Mac Studio)",,2x Apple M5 Ultra,,,Mistral Medium 3.5 128B,"FP8, 8,192-token context","MLX, 2-node tensor parallel",13.69,none,,,,,yes,https://machinelearning.apple.com/research/mlx-distributed-inference,https://getrigbase.com/builds/2x-mac-studio-m5-ultra-rdma-cluster/
2x-mac-studio-m5-ultra-rdma-cluster,2x Mac Studio M5 Ultra cluster (Thunderbolt 5 RDMA): Muse Spark Open 816B,"2x Apple M5 Ultra (unified, Mac Studio)",,2x Apple M5 Ultra,,,Qwen3.8 27B,"BF16, 8,192-token context","MLX, 2-node tensor parallel",33.27,none,,,,,yes,https://machinelearning.apple.com/research/mlx-distributed-inference,https://getrigbase.com/builds/2x-mac-studio-m5-ultra-rdma-cluster/
4x-mac-studio-m5-ultra-rdma-cluster,4x Mac Studio M5 Ultra cluster (Thunderbolt 5 RDMA): Muse Spark Open at 36.7 tok/s,"4x Apple M5 Ultra (unified, Mac Studio)",,4x Apple M5 Ultra,,,Muse Spark Open 816B-A42B,"NVFP4, 8,192-token context","MLX, 4-node tensor parallel",36.73,none,,,,,yes,https://machinelearning.apple.com/research/mlx-distributed-inference,https://getrigbase.com/builds/4x-mac-studio-m5-ultra-rdma-cluster/
4x-mac-studio-m5-ultra-rdma-cluster,4x Mac Studio M5 Ultra cluster (Thunderbolt 5 RDMA): Muse Spark Open at 36.7 tok/s,"4x Apple M5 Ultra (unified, Mac Studio)",,4x Apple M5 Ultra,,,Mistral Medium 3.5 128B,"FP8, 8,192-token context","MLX, 4-node tensor parallel",23.37,none,,,,,yes,https://machinelearning.apple.com/research/mlx-distributed-inference,https://getrigbase.com/builds/4x-mac-studio-m5-ultra-rdma-cluster/
4x-mac-studio-m5-ultra-rdma-cluster,4x Mac Studio M5 Ultra cluster (Thunderbolt 5 RDMA): Muse Spark Open at 36.7 tok/s,"4x Apple M5 Ultra (unified, Mac Studio)",,4x Apple M5 Ultra,,,Qwen3.8 27B,"BF16, 8,192-token context","MLX, 4-node tensor parallel",50.89,none,,,,,yes,https://machinelearning.apple.com/research/mlx-distributed-inference,https://getrigbase.com/builds/4x-mac-studio-m5-ultra-rdma-cluster/
rtx-3090-qwen38-27b-openjet-megakernel,Single RTX 3090: Qwen3.8 27B at 78-140 tok/s with the OpenJet megakernel vs llama.cpp,RTX 3090,24,,,,Qwen3.8 27B,"Q4_K_M (unsloth), greedy code prompt","OpenJet megakernel (MTP, 4 drafts)",139.7,MTP,,,,,yes,https://github.com/L-Forster/open-jet/tree/master/megakernel,https://getrigbase.com/builds/rtx-3090-qwen38-27b-openjet-megakernel/
rtx-3090-qwen38-27b-openjet-megakernel,Single RTX 3090: Qwen3.8 27B at 78-140 tok/s with the OpenJet megakernel vs llama.cpp,RTX 3090,24,,,,Qwen3.8 27B,"Q4_K_M (unsloth), greedy code prompt","llama.cpp (MTP, 2 drafts)",72.5,MTP,,,,,yes,https://github.com/L-Forster/open-jet/tree/master/megakernel,https://getrigbase.com/builds/rtx-3090-qwen38-27b-openjet-megakernel/
rtx-3090-qwen38-27b-openjet-megakernel,Single RTX 3090: Qwen3.8 27B at 78-140 tok/s with the OpenJet megakernel vs llama.cpp,RTX 3090,24,,,,Qwen3.8 27B,"Q4_K_M (unsloth), reasoning mode",OpenJet megakernel (MTP),78.0,MTP,,,,,yes,https://github.com/L-Forster/open-jet/tree/master/megakernel,https://getrigbase.com/builds/rtx-3090-qwen38-27b-openjet-megakernel/
rtx-3090-qwen38-27b-openjet-megakernel,Single RTX 3090: Qwen3.8 27B at 78-140 tok/s with the OpenJet megakernel vs llama.cpp,RTX 3090,24,,,,Qwen3.8 27B,"Q4_K_M (unsloth), reasoning mode","llama.cpp (MTP, 2 drafts)",50.9,MTP,,,,,yes,https://github.com/L-Forster/open-jet/tree/master/megakernel,https://getrigbase.com/builds/rtx-3090-qwen38-27b-openjet-megakernel/
rtx-3090-qwen38-27b-openjet-megakernel,Single RTX 3090: Qwen3.8 27B at 78-140 tok/s with the OpenJet megakernel vs llama.cpp,RTX 3090,24,,,,Qwen3.8 27B,"Q4_K_M (unsloth), no speculation",llama.cpp,40.4,none,,,,,yes,https://github.com/L-Forster/open-jet/tree/master/megakernel,https://getrigbase.com/builds/rtx-3090-qwen38-27b-openjet-megakernel/
rtx-3090-qwen38-27b-openjet-megakernel,Single RTX 3090: Qwen3.8 27B at 78-140 tok/s with the OpenJet megakernel vs llama.cpp,RTX 3090,24,,,,Qwen3.8 27B,"Q4_K_M (unsloth), no speculation",OpenJet megakernel,42.9,none,,,,,yes,https://github.com/L-Forster/open-jet/tree/master/megakernel,https://getrigbase.com/builds/rtx-3090-qwen38-27b-openjet-megakernel/
hp-omnibook-ultra-16-rtx-spark-128gb,"HP OmniBook Ultra 16 (RTX Spark N1X, 128GB) laptop: local AI benchmarks","Nvidia RTX Spark N1X, 6,144-core Blackwell GPU (unified)",128,"Nvidia RTX Spark N1X, 20-core Grace CPU","128GB LPDDR5X-9400 (64GB GPU / 64GB system split, 62.5GB GPU-usable)",,Qwen3.6 35B-A3B,"Ollama default, 64K context, Codex CLI coding agent",Ollama,89.5,none,,,,,yes,https://www.storagereview.com/review/hp-omnibook-ultra-16-review-rtx-spark-n1x,https://getrigbase.com/builds/hp-omnibook-ultra-16-rtx-spark-128gb/
hp-omnibook-ultra-16-rtx-spark-128gb,"HP OmniBook Ultra 16 (RTX Spark N1X, 128GB) laptop: local AI benchmarks","Nvidia RTX Spark N1X, 6,144-core Blackwell GPU (unified)",128,"Nvidia RTX Spark N1X, 20-core Grace CPU","128GB LPDDR5X-9400 (64GB GPU / 64GB system split, 62.5GB GPU-usable)",,gpt-oss 20B,"Ollama default, 64K context, Codex CLI coding agent",Ollama,58.0,none,,,,,yes,https://www.storagereview.com/review/hp-omnibook-ultra-16-review-rtx-spark-n1x,https://getrigbase.com/builds/hp-omnibook-ultra-16-rtx-spark-128gb/
hp-omnibook-ultra-16-rtx-spark-128gb,"HP OmniBook Ultra 16 (RTX Spark N1X, 128GB) laptop: local AI benchmarks","Nvidia RTX Spark N1X, 6,144-core Blackwell GPU (unified)",128,"Nvidia RTX Spark N1X, 20-core Grace CPU","128GB LPDDR5X-9400 (64GB GPU / 64GB system split, 62.5GB GPU-usable)",,Llama 3.1 8B,MLPerf Client v2.0 (geomean),llama.cpp CUDA,56.5,none,,,,,yes,https://www.storagereview.com/review/hp-omnibook-ultra-16-review-rtx-spark-n1x,https://getrigbase.com/builds/hp-omnibook-ultra-16-rtx-spark-128gb/
surface-laptop-ultra-rtx-spark-microcenter,Microsoft Surface Laptop Ultra (RTX Spark N1X): llama.cpp speeds on 2026 models (Micro Center),"Nvidia RTX Spark N1X, 6,144-core Blackwell GPU (unified)",,"Nvidia RTX Spark N1X, 20-core Grace CPU",,Avg GPU power while generating: 43-48 W,Nemotron 3 Nano Omni 30B,"general, short (~90-token) prompt",llama.cpp (Nvidia ARM64 build),63.9,none,,,,,yes,https://www.microcenter.com/site/mc-news/article/benchmarking-the-rtx-spark.aspx,https://getrigbase.com/builds/surface-laptop-ultra-rtx-spark-microcenter/
surface-laptop-ultra-rtx-spark-microcenter,Microsoft Surface Laptop Ultra (RTX Spark N1X): llama.cpp speeds on 2026 models (Micro Center),"Nvidia RTX Spark N1X, 6,144-core Blackwell GPU (unified)",,"Nvidia RTX Spark N1X, 20-core Grace CPU",,Avg GPU power while generating: 43-48 W,GLM 4.7 Flash,"general, short prompt",llama.cpp (Nvidia ARM64 build),55.6,none,,,,,yes,https://www.microcenter.com/site/mc-news/article/benchmarking-the-rtx-spark.aspx,https://getrigbase.com/builds/surface-laptop-ultra-rtx-spark-microcenter/
surface-laptop-ultra-rtx-spark-microcenter,Microsoft Surface Laptop Ultra (RTX Spark N1X): llama.cpp speeds on 2026 models (Micro Center),"Nvidia RTX Spark N1X, 6,144-core Blackwell GPU (unified)",,"Nvidia RTX Spark N1X, 20-core Grace CPU",,Avg GPU power while generating: 43-48 W,Qwen3.5 122B,"A10B, general, short prompt",llama.cpp (Nvidia ARM64 build),35.2,none,,,,,yes,https://www.microcenter.com/site/mc-news/article/benchmarking-the-rtx-spark.aspx,https://getrigbase.com/builds/surface-laptop-ultra-rtx-spark-microcenter/
surface-laptop-ultra-rtx-spark-microcenter,Microsoft Surface Laptop Ultra (RTX Spark N1X): llama.cpp speeds on 2026 models (Micro Center),"Nvidia RTX Spark N1X, 6,144-core Blackwell GPU (unified)",,"Nvidia RTX Spark N1X, 20-core Grace CPU",,Avg GPU power while generating: 43-48 W,Muse Glimmer 30B,"general, short prompt",llama.cpp (Nvidia ARM64 build),15.9,none,,,,,yes,https://www.microcenter.com/site/mc-news/article/benchmarking-the-rtx-spark.aspx,https://getrigbase.com/builds/surface-laptop-ultra-rtx-spark-microcenter/
surface-laptop-ultra-rtx-spark-microcenter,Microsoft Surface Laptop Ultra (RTX Spark N1X): llama.cpp speeds on 2026 models (Micro Center),"Nvidia RTX Spark N1X, 6,144-core Blackwell GPU (unified)",,"Nvidia RTX Spark N1X, 20-core Grace CPU",,Avg GPU power while generating: 43-48 W,Qwen3.8 27B,"general, short prompt",llama.cpp (Nvidia ARM64 build),14.5,none,,,,,yes,https://www.microcenter.com/site/mc-news/article/benchmarking-the-rtx-spark.aspx,https://getrigbase.com/builds/surface-laptop-ultra-rtx-spark-microcenter/
dgx-spark-nemotron-35-super-vl-bolt,"DGX Spark (GB10, 128GB): Nemotron 3.5 Super VL 120B and Bolt 74B, launch-week runs",Nvidia GB10 (unified),128,Nvidia GB10,128GB unified memory,GPU peak 82 W (nvidia-smi) during Super VL runs with prompts up to 64K tokens,Nemotron 3.5 Super VL 120B-A12B,"NVFP4, no speculative decoding, 256-token output",vLLM,19.06,none,,,,,yes,https://dev.classmethod.jp/articles/dgx-spark-nemotron-3-5-super-vl-first-touch/,https://getrigbase.com/builds/dgx-spark-nemotron-35-super-vl-bolt/
dgx-spark-nemotron-35-super-vl-bolt,"DGX Spark (GB10, 128GB): Nemotron 3.5 Super VL 120B and Bolt 74B, launch-week runs",Nvidia GB10 (unified),128,Nvidia GB10,128GB unified memory,GPU peak 82 W (nvidia-smi) during Super VL runs with prompts up to 64K tokens,Nemotron 3.5 Super VL 120B-A12B,NVFP4 + LiLiCorr speculative decoding (7 draft tokens),vLLM,38.07,LiLiCorr,,,,,yes,https://dev.classmethod.jp/articles/dgx-spark-nemotron-3-5-super-vl-first-touch/,https://getrigbase.com/builds/dgx-spark-nemotron-35-super-vl-bolt/
dgx-spark-nemotron-35-super-vl-bolt,"DGX Spark (GB10, 128GB): Nemotron 3.5 Super VL 120B and Bolt 74B, launch-week runs",Nvidia GB10 (unified),128,Nvidia GB10,128GB unified memory,GPU peak 82 W (nvidia-smi) during Super VL runs with prompts up to 64K tokens,Nemotron 3.5 Bolt 74B-A9B,"2-bit 24GB GGUF, no speculative decoding",llama.cpp,39.47,none,,,,,yes,https://dev.classmethod.jp/articles/dgx-spark-nemotron-3-5-super-vl-first-touch/,https://getrigbase.com/builds/dgx-spark-nemotron-35-super-vl-bolt/
dgx-spark-nemotron-35-super-vl-bolt,"DGX Spark (GB10, 128GB): Nemotron 3.5 Super VL 120B and Bolt 74B, launch-week runs",Nvidia GB10 (unified),128,Nvidia GB10,128GB unified memory,GPU peak 82 W (nvidia-smi) during Super VL runs with prompts up to 64K tokens,Nemotron 3.5 Bolt 74B-A9B,2-bit 24GB GGUF + DFlash speculative decoding (7 draft tokens),llama.cpp,61.61,DFlash,,,,,yes,https://dev.classmethod.jp/articles/dgx-spark-nemotron-3-5-super-vl-first-touch/,https://getrigbase.com/builds/dgx-spark-nemotron-35-super-vl-bolt/
r9700-qwen36-35b-a3b-amdai,Radeon AI Pro R9700 (one card): Qwen3.6 35B-A3B from 4K to 256K context (AMDai lab rig),AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,256 GB DDR5,,Qwen3.6 35B-A3B,"Unsloth UD-Q5_K_XL, short context, f16 KV",llama.cpp Vulkan,102.39,none,,,,,yes,https://amdai.net/,https://getrigbase.com/builds/r9700-qwen36-35b-a3b-amdai/
r9700-qwen36-35b-a3b-amdai,Radeon AI Pro R9700 (one card): Qwen3.6 35B-A3B from 4K to 256K context (AMDai lab rig),AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,256 GB DDR5,,Qwen3.6 35B-A3B,"UD-Q5_K_XL, 32K context",llama.cpp Vulkan,90.52,none,,,,,yes,https://amdai.net/,https://getrigbase.com/builds/r9700-qwen36-35b-a3b-amdai/
r9700-qwen36-35b-a3b-amdai,Radeon AI Pro R9700 (one card): Qwen3.6 35B-A3B from 4K to 256K context (AMDai lab rig),AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,256 GB DDR5,,Qwen3.6 35B-A3B,"UD-Q5_K_XL, 128K context",llama.cpp Vulkan,66.43,none,,,,,yes,https://amdai.net/,https://getrigbase.com/builds/r9700-qwen36-35b-a3b-amdai/
r9700-qwen36-35b-a3b-amdai,Radeon AI Pro R9700 (one card): Qwen3.6 35B-A3B from 4K to 256K context (AMDai lab rig),AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,256 GB DDR5,,Qwen3.6 35B-A3B,"UD-Q5_K_XL, 256K context",llama.cpp Vulkan,44.4,none,,,,,yes,https://amdai.net/,https://getrigbase.com/builds/r9700-qwen36-35b-a3b-amdai/
r9700-qwen36-35b-a3b-amdai,Radeon AI Pro R9700 (one card): Qwen3.6 35B-A3B from 4K to 256K context (AMDai lab rig),AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,256 GB DDR5,,Qwen3.6 35B-A3B,Q5_K_M (24.68 GB VRAM),llama.cpp Vulkan,109.73,none,,,,,yes,https://amdai.net/,https://getrigbase.com/builds/r9700-qwen36-35b-a3b-amdai/
r9700-qwen36-35b-a3b-amdai,Radeon AI Pro R9700 (one card): Qwen3.6 35B-A3B from 4K to 256K context (AMDai lab rig),AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,256 GB DDR5,,Qwen3.6 35B-A3B,Q4_K_M,llama.cpp Vulkan,117.69,none,,,,,yes,https://amdai.net/,https://getrigbase.com/builds/r9700-qwen36-35b-a3b-amdai/
r9700-qwen36-27b-mtp-backends,"Radeon AI Pro R9700: Qwen3.6 27B on Vulkan, HIP and Ollama, with and without MTP",AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,,,Qwen3.6 27B,"Q4_K_M + MTP draft head (draft-n-max 3), 64K context",llama.cpp Vulkan,66.0,MTP,,,,,yes,https://localaifrontier.com/benchmarks/qwen3-6-27b-mtp-backend-showdown-r9700/,https://getrigbase.com/builds/r9700-qwen36-27b-mtp-backends/
r9700-qwen36-27b-mtp-backends,"Radeon AI Pro R9700: Qwen3.6 27B on Vulkan, HIP and Ollama, with and without MTP",AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,,,Qwen3.6 27B,"Q4_K_M, no MTP",llama.cpp Vulkan,31.9,none,,,,,yes,https://localaifrontier.com/benchmarks/qwen3-6-27b-mtp-backend-showdown-r9700/,https://getrigbase.com/builds/r9700-qwen36-27b-mtp-backends/
r9700-qwen36-27b-mtp-backends,"Radeon AI Pro R9700: Qwen3.6 27B on Vulkan, HIP and Ollama, with and without MTP",AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,,,Qwen3.6 27B,Q4_K_M,llama.cpp HIP (ROCm),27.1,none,,,,,yes,https://localaifrontier.com/benchmarks/qwen3-6-27b-mtp-backend-showdown-r9700/,https://getrigbase.com/builds/r9700-qwen36-27b-mtp-backends/
r9700-qwen36-27b-mtp-backends,"Radeon AI Pro R9700: Qwen3.6 27B on Vulkan, HIP and Ollama, with and without MTP",AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,,,Qwen3.6 27B,Q4_K_M,Ollama (HIP),26.7,none,,,,,yes,https://localaifrontier.com/benchmarks/qwen3-6-27b-mtp-backend-showdown-r9700/,https://getrigbase.com/builds/r9700-qwen36-27b-mtp-backends/
r9700-qwen38-27b-vaditaslim,"Radeon AI Pro R9700: Qwen3.8 27B on llama.cpp, Ollama and vLLM (Vadi Taslim)",AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,,,Qwen3.8 27B,"Q4_K_M + MTP speculative decoding, reasoning on",llama.cpp Vulkan,57,MTP,,,,,yes,https://www.vaditaslim.com/blog/ai/qwen3.8-27b-two-rigs,https://getrigbase.com/builds/r9700-qwen38-27b-vaditaslim/
r9700-qwen38-27b-vaditaslim,"Radeon AI Pro R9700: Qwen3.8 27B on llama.cpp, Ollama and vLLM (Vadi Taslim)",AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,,,Qwen3.8 27B,Q4_K_M,Ollama,30,none,,,,,yes,https://www.vaditaslim.com/blog/ai/qwen3.8-27b-two-rigs,https://getrigbase.com/builds/r9700-qwen38-27b-vaditaslim/
r9700-qwen38-27b-vaditaslim,"Radeon AI Pro R9700: Qwen3.8 27B on llama.cpp, Ollama and vLLM (Vadi Taslim)",AMD Radeon AI Pro R9700 (1 of 2 cards in the rig used),32,,,,Qwen3.8 27B,"AWQ INT4, single request",vLLM (radiance),22,none,,,,,yes,https://www.vaditaslim.com/blog/ai/qwen3.8-27b-two-rigs,https://getrigbase.com/builds/r9700-qwen38-27b-vaditaslim/
rtx-pro-6000-maxq-qwen38-27b-vaditaslim,"RTX Pro 6000 Blackwell Max-Q (300 W cap): Qwen3.8 27B on vLLM, NVFP4 vs BF16","RTX Pro 6000 Max-Q (Blackwell, 96 GB)",96,,,GPU power cap 300 W,Qwen3.8 27B,"NVFP4 (W4A16) + MTP (5 draft tokens), 256K context",vLLM 0.27.1,150,MTP,,,,,yes,https://www.vaditaslim.com/blog/ai/qwen3.8-27b-two-rigs,https://getrigbase.com/builds/rtx-pro-6000-maxq-qwen38-27b-vaditaslim/
rtx-pro-6000-maxq-qwen38-27b-vaditaslim,"RTX Pro 6000 Blackwell Max-Q (300 W cap): Qwen3.8 27B on vLLM, NVFP4 vs BF16","RTX Pro 6000 Max-Q (Blackwell, 96 GB)",96,,,GPU power cap 300 W,Qwen3.8 27B,"BF16 + MTP (5 draft tokens), 256K context",vLLM 0.20.1,85,MTP,,,,,yes,https://www.vaditaslim.com/blog/ai/qwen3.8-27b-two-rigs,https://getrigbase.com/builds/rtx-pro-6000-maxq-qwen38-27b-vaditaslim/
rx-7900-xtx-qwen36-35b-a3b-mtp,"Radeon RX 7900 XTX (24 GB): Qwen3.6 35B-A3B up to 245K context, with MTP",AMD Radeon RX 7900 XTX,24,AMD Ryzen 5 3600,128 GiB,,Qwen3.6 35B-A3B,"UD-Q4_K_XL, llama-bench 2K, f16 KV",llama.cpp ROCm,87.2,none,,,,,yes,https://gist.github.com/AshyIsMe/ef9dc0e57d7af5626f3935f55c38543a,https://getrigbase.com/builds/rx-7900-xtx-qwen36-35b-a3b-mtp/
rx-7900-xtx-qwen36-35b-a3b-mtp,"Radeon RX 7900 XTX (24 GB): Qwen3.6 35B-A3B up to 245K context, with MTP",AMD Radeon RX 7900 XTX,24,AMD Ryzen 5 3600,128 GiB,,Qwen3.6 35B-A3B,"UD-Q4_K_XL, ~16K real-text context, q4_0 KV",llama.cpp ROCm (llama-server),79.6,none,,,,,yes,https://gist.github.com/AshyIsMe/ef9dc0e57d7af5626f3935f55c38543a,https://getrigbase.com/builds/rx-7900-xtx-qwen36-35b-a3b-mtp/
rx-7900-xtx-qwen36-35b-a3b-mtp,"Radeon RX 7900 XTX (24 GB): Qwen3.6 35B-A3B up to 245K context, with MTP",AMD Radeon RX 7900 XTX,24,AMD Ryzen 5 3600,128 GiB,,Qwen3.6 35B-A3B,"UD-Q4_K_XL + MTP (draft-n-max 4), ~16K context",llama.cpp ROCm (llama-server),135.2,MTP,,,,,yes,https://gist.github.com/AshyIsMe/ef9dc0e57d7af5626f3935f55c38543a,https://getrigbase.com/builds/rx-7900-xtx-qwen36-35b-a3b-mtp/
rx-7900-xtx-qwen36-35b-a3b-mtp,"Radeon RX 7900 XTX (24 GB): Qwen3.6 35B-A3B up to 245K context, with MTP",AMD Radeon RX 7900 XTX,24,AMD Ryzen 5 3600,128 GiB,,Qwen3.6 35B-A3B,"UD-Q4_K_XL, ~64K context",llama.cpp ROCm (llama-server),66.0,none,,,,,yes,https://gist.github.com/AshyIsMe/ef9dc0e57d7af5626f3935f55c38543a,https://getrigbase.com/builds/rx-7900-xtx-qwen36-35b-a3b-mtp/
rx-7900-xtx-qwen36-35b-a3b-mtp,"Radeon RX 7900 XTX (24 GB): Qwen3.6 35B-A3B up to 245K context, with MTP",AMD Radeon RX 7900 XTX,24,AMD Ryzen 5 3600,128 GiB,,Qwen3.6 35B-A3B,"UD-Q4_K_XL + MTP, ~64K context",llama.cpp ROCm (llama-server),108.2,MTP,,,,,yes,https://gist.github.com/AshyIsMe/ef9dc0e57d7af5626f3935f55c38543a,https://getrigbase.com/builds/rx-7900-xtx-qwen36-35b-a3b-mtp/
rx-7900-xtx-qwen36-35b-a3b-mtp,"Radeon RX 7900 XTX (24 GB): Qwen3.6 35B-A3B up to 245K context, with MTP",AMD Radeon RX 7900 XTX,24,AMD Ryzen 5 3600,128 GiB,,Qwen3.6 35B-A3B,"UD-Q4_K_XL, ~245K context, q4_0 KV",llama.cpp ROCm (llama-server),38.3,none,,,,,yes,https://gist.github.com/AshyIsMe/ef9dc0e57d7af5626f3935f55c38543a,https://getrigbase.com/builds/rx-7900-xtx-qwen36-35b-a3b-mtp/
strix-halo-corsair-ai-300-rocm-vs-vulkan,"Strix Halo (Corsair AI Workstation 300, 128 GB): ROCm vs Vulkan on Qwen3.8 27B and Qwen3.6 35B-A3B",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,"128 GiB (8 x 16 GB, 'LPDDR5' per dmidecode, 8000 MT/s configured)",Balanced power profile,Qwen3.6 35B-A3B,"Q8_0, llama-bench tg128",llama.cpp Vulkan,54.22,none,,,,,yes,https://howquant.dev/benchmarks/strix-halo-rocm-vs-vulkan/,https://getrigbase.com/builds/strix-halo-corsair-ai-300-rocm-vs-vulkan/
strix-halo-corsair-ai-300-rocm-vs-vulkan,"Strix Halo (Corsair AI Workstation 300, 128 GB): ROCm vs Vulkan on Qwen3.8 27B and Qwen3.6 35B-A3B",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,"128 GiB (8 x 16 GB, 'LPDDR5' per dmidecode, 8000 MT/s configured)",Balanced power profile,Qwen3.6 35B-A3B,"Q8_0, llama-bench tg128",llama.cpp ROCm,47.47,none,,,,,yes,https://howquant.dev/benchmarks/strix-halo-rocm-vs-vulkan/,https://getrigbase.com/builds/strix-halo-corsair-ai-300-rocm-vs-vulkan/
strix-halo-corsair-ai-300-rocm-vs-vulkan,"Strix Halo (Corsair AI Workstation 300, 128 GB): ROCm vs Vulkan on Qwen3.8 27B and Qwen3.6 35B-A3B",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,"128 GiB (8 x 16 GB, 'LPDDR5' per dmidecode, 8000 MT/s configured)",Balanced power profile,Qwen3.6 35B-A3B,"Q8_0 + MTP (draft-n-max 3), coding workload",llama.cpp ROCm,60.92,MTP,,,,,yes,https://howquant.dev/benchmarks/strix-halo-rocm-vs-vulkan/,https://getrigbase.com/builds/strix-halo-corsair-ai-300-rocm-vs-vulkan/
strix-halo-corsair-ai-300-rocm-vs-vulkan,"Strix Halo (Corsair AI Workstation 300, 128 GB): ROCm vs Vulkan on Qwen3.8 27B and Qwen3.6 35B-A3B",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,"128 GiB (8 x 16 GB, 'LPDDR5' per dmidecode, 8000 MT/s configured)",Balanced power profile,Qwen3.8 27B,"Q8_0, llama-bench tg128",llama.cpp ROCm,7.76,none,,,,,yes,https://howquant.dev/benchmarks/strix-halo-rocm-vs-vulkan/,https://getrigbase.com/builds/strix-halo-corsair-ai-300-rocm-vs-vulkan/
strix-halo-corsair-ai-300-rocm-vs-vulkan,"Strix Halo (Corsair AI Workstation 300, 128 GB): ROCm vs Vulkan on Qwen3.8 27B and Qwen3.6 35B-A3B",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,"128 GiB (8 x 16 GB, 'LPDDR5' per dmidecode, 8000 MT/s configured)",Balanced power profile,Qwen3.8 27B,"Q8_0, llama-bench tg128",llama.cpp Vulkan,7.77,none,,,,,yes,https://howquant.dev/benchmarks/strix-halo-rocm-vs-vulkan/,https://getrigbase.com/builds/strix-halo-corsair-ai-300-rocm-vs-vulkan/
strix-halo-corsair-ai-300-rocm-vs-vulkan,"Strix Halo (Corsair AI Workstation 300, 128 GB): ROCm vs Vulkan on Qwen3.8 27B and Qwen3.6 35B-A3B",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,"128 GiB (8 x 16 GB, 'LPDDR5' per dmidecode, 8000 MT/s configured)",Balanced power profile,Qwen3.8 27B,"Q8_0 + MTP (draft-n-max 3), coding workload, reasoning on",llama.cpp ROCm,15.67,MTP,,,,,yes,https://howquant.dev/benchmarks/strix-halo-rocm-vs-vulkan/,https://getrigbase.com/builds/strix-halo-corsair-ai-300-rocm-vs-vulkan/
strix-halo-corsair-ai-300-rocm-vs-vulkan,"Strix Halo (Corsair AI Workstation 300, 128 GB): ROCm vs Vulkan on Qwen3.8 27B and Qwen3.6 35B-A3B",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,"128 GiB (8 x 16 GB, 'LPDDR5' per dmidecode, 8000 MT/s configured)",Balanced power profile,Qwen3.8 27B,"Q8_0 + MTP (draft-n-max 3), coding workload, reasoning on",llama.cpp Vulkan,13.19,MTP,,,,,yes,https://howquant.dev/benchmarks/strix-halo-rocm-vs-vulkan/,https://getrigbase.com/builds/strix-halo-corsair-ai-300-rocm-vs-vulkan/
strix-halo-qwen38-27b-mtp-sleepingrobots,"Strix Halo (Ryzen AI Max+ 395, 128 GB): Qwen3.8 27B Q6_K, 9.6 to 18 tok/s with MTP",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,128 GB LPDDR5X unified memory,,Qwen3.8 27B,"Q6_K, no MTP",llama.cpp Vulkan (RADV),9.59,none,,,,,yes,https://sleepingrobots.com/dreams/qwen38-mtp-strix-halo/,https://getrigbase.com/builds/strix-halo-qwen38-27b-mtp-sleepingrobots/
strix-halo-qwen38-27b-mtp-sleepingrobots,"Strix Halo (Ryzen AI Max+ 395, 128 GB): Qwen3.8 27B Q6_K, 9.6 to 18 tok/s with MTP",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,128 GB LPDDR5X unified memory,,Qwen3.8 27B,Q6_K + MTP (draft-n-max 2),llama.cpp Vulkan (RADV),18.01,MTP,,,,,yes,https://sleepingrobots.com/dreams/qwen38-mtp-strix-halo/,https://getrigbase.com/builds/strix-halo-qwen38-27b-mtp-sleepingrobots/
strix-halo-qwen38-27b-mtp-sleepingrobots,"Strix Halo (Ryzen AI Max+ 395, 128 GB): Qwen3.8 27B Q6_K, 9.6 to 18 tok/s with MTP",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,128 GB LPDDR5X unified memory,,Qwen3.8 27B,Q6_K + MTP (draft-n-max 5),llama.cpp Vulkan (RADV),15.16,MTP,,,,,yes,https://sleepingrobots.com/dreams/qwen38-mtp-strix-halo/,https://getrigbase.com/builds/strix-halo-qwen38-27b-mtp-sleepingrobots/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Qwen3.6 35B-A3B,NVFP4 + MTP (k=2),"vLLM 0.31.0, 128K ctx, 1 user",102.3,MTP,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Qwen3.6 35B-A3B,"NVFP4, no speculative decoding","vLLM 0.31.0, 128K ctx, 1 user",78.1,none,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,gpt-oss 20B,,"vLLM 0.31.0, 128K ctx, 1 user",96.2,none,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Nemotron 3.5 Lightning 30B-A3B,+ DSpark speculative decoding (k=3),"vLLM 0.31.0, 128K ctx, 1 user",94.4,DSpark,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Nemotron 3.5 Lightning 30B-A3B,no speculative decoding,"vLLM 0.31.0, 128K ctx, 1 user",70.3,none,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Nemotron 3 Nano Omni 30B,,"vLLM 0.31.0, 128K ctx, 1 user",58.8,none,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Gemma 4 26B-A4B,+ official MTP drafter (k=3),"vLLM 0.31.0, 128K ctx, 1 user",56.5,MTP,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Gemma 4 26B-A4B,no speculative decoding,"vLLM 0.31.0, 128K ctx, 1 user",30.8,none,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,gpt-oss 120B,,"vLLM 0.31.0, 128K ctx, 1 user",44.4,none,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Qwen3.8 Flash-Next,+ MTP (k=2),"patched vLLM (qwen38-flash-dgx), 128K ctx, 1 user",26.8,MTP,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Nemotron 3 Super 120B-A12B,+ MTP (k=3),"vLLM 0.31.0, 128K ctx, 1 user",23.0,MTP,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Nemotron 3 Super 120B-A12B,no speculative decoding,"vLLM 0.31.0, 128K ctx, 1 user",16.1,none,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Qwen3.8 27B,+ MTP (k=2),"vLLM 0.31.0, 128K ctx, 1 user",19.0,MTP,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Qwen3.8 27B,no speculative decoding,"vLLM 0.31.0, 128K ctx, 1 user",11.0,none,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
dgx-spark-vllm-23-models-ctala-oct-2026,"DGX Spark (GB10, 128 GB): 23 models on vLLM at 128K context, Qwen3.6 35B-A3B up to 102 tok/s",Nvidia GB10 (unified),128,Nvidia GB10 Grace (Arm),128 GB LPDDR5X unified (~121 GB usable),,Llama 3.1 8B,,"NVIDIA NIM 2.0.9, 128K ctx, 1 user",26.4,none,,,,,yes,https://ctala.github.io/local-llm-agentic-workflows/benchmarks/,https://getrigbase.com/builds/dgx-spark-vllm-23-models-ctala-oct-2026/
minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb,"Minisforum MS-S1 MAX (Ryzen AI Max+ PRO 495, 192 GB): gpt-oss 120B at 55.8 tok/s","AMD Ryzen AI Max+ PRO 495 (Radeon 8065S iGPU, 40 CUs)",192,"AMD Ryzen AI Max+ PRO 495 (""Gorgon Halo"")",192 GB LPDDR5X-8533 unified (GPU limit raised to 170 GB via ttm.pages_limit),,gpt-oss 120B,MXFP4,llama.cpp Vulkan,55.8,none,,,,,yes,https://www.hommie.net/posts/ai-box.html,https://getrigbase.com/builds/minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb/
minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb,"Minisforum MS-S1 MAX (Ryzen AI Max+ PRO 495, 192 GB): gpt-oss 120B at 55.8 tok/s","AMD Ryzen AI Max+ PRO 495 (Radeon 8065S iGPU, 40 CUs)",192,"AMD Ryzen AI Max+ PRO 495 (""Gorgon Halo"")",192 GB LPDDR5X-8533 unified (GPU limit raised to 170 GB via ttm.pages_limit),,gpt-oss 120B,"MXFP4, 32K context",llama.cpp Vulkan,45.1,none,,,,,yes,https://www.hommie.net/posts/ai-box.html,https://getrigbase.com/builds/minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb/
minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb,"Minisforum MS-S1 MAX (Ryzen AI Max+ PRO 495, 192 GB): gpt-oss 120B at 55.8 tok/s","AMD Ryzen AI Max+ PRO 495 (Radeon 8065S iGPU, 40 CUs)",192,"AMD Ryzen AI Max+ PRO 495 (""Gorgon Halo"")",192 GB LPDDR5X-8533 unified (GPU limit raised to 170 GB via ttm.pages_limit),,MiniMax M2.7,IQ4_XS,llama.cpp Vulkan,29.5,none,,,,,yes,https://www.hommie.net/posts/ai-box.html,https://getrigbase.com/builds/minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb/
minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb,"Minisforum MS-S1 MAX (Ryzen AI Max+ PRO 495, 192 GB): gpt-oss 120B at 55.8 tok/s","AMD Ryzen AI Max+ PRO 495 (Radeon 8065S iGPU, 40 CUs)",192,"AMD Ryzen AI Max+ PRO 495 (""Gorgon Halo"")",192 GB LPDDR5X-8533 unified (GPU limit raised to 170 GB via ttm.pages_limit),,MiniMax M2.7,"IQ4_XS, 32K context",llama.cpp Vulkan,17.3,none,,,,,yes,https://www.hommie.net/posts/ai-box.html,https://getrigbase.com/builds/minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb/
minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb,"Minisforum MS-S1 MAX (Ryzen AI Max+ PRO 495, 192 GB): gpt-oss 120B at 55.8 tok/s","AMD Ryzen AI Max+ PRO 495 (Radeon 8065S iGPU, 40 CUs)",192,"AMD Ryzen AI Max+ PRO 495 (""Gorgon Halo"")",192 GB LPDDR5X-8533 unified (GPU limit raised to 170 GB via ttm.pages_limit),,DeepSeek V4 Flash,IQ3_S,llama.cpp Vulkan,17.4,none,,,,,yes,https://www.hommie.net/posts/ai-box.html,https://getrigbase.com/builds/minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb/
minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb,"Minisforum MS-S1 MAX (Ryzen AI Max+ PRO 495, 192 GB): gpt-oss 120B at 55.8 tok/s","AMD Ryzen AI Max+ PRO 495 (Radeon 8065S iGPU, 40 CUs)",192,"AMD Ryzen AI Max+ PRO 495 (""Gorgon Halo"")",192 GB LPDDR5X-8533 unified (GPU limit raised to 170 GB via ttm.pages_limit),,DeepSeek V4 Flash,"IQ3_S, 32K context",llama.cpp Vulkan,12.2,none,,,,,yes,https://www.hommie.net/posts/ai-box.html,https://getrigbase.com/builds/minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb/
minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb,"Minisforum MS-S1 MAX (Ryzen AI Max+ PRO 495, 192 GB): gpt-oss 120B at 55.8 tok/s","AMD Ryzen AI Max+ PRO 495 (Radeon 8065S iGPU, 40 CUs)",192,"AMD Ryzen AI Max+ PRO 495 (""Gorgon Halo"")",192 GB LPDDR5X-8533 unified (GPU limit raised to 170 GB via ttm.pages_limit),,Qwen3.8 27B,"Q8 (author: ""about 8 tok/s"")",llama.cpp Vulkan,8,none,,,,,yes,https://www.hommie.net/posts/ai-box.html,https://getrigbase.com/builds/minisforum-ms-s1-max-ryzen-ai-max-pro-495-192gb/
rtx-5070-12gb-qwen38-flash-next-expert-streaming,RTX 5070 12GB + 32 GB DDR4: Qwen3.8-Flash-Next 177B at 11.5 tok/s with expert streaming,RTX 5070 12GB,12,AMD Ryzen 5 5600GT,32 GB DDR4-2400,,Qwen3.8 Flash-Next,"UD-IQ3_XXS, experts streamed from RAM/SSD",llama.cpp-based expert streaming (custom),11.5,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/29930,https://getrigbase.com/builds/rtx-5070-12gb-qwen38-flash-next-expert-streaming/
rtx-5070-12gb-qwen38-flash-next-expert-streaming,RTX 5070 12GB + 32 GB DDR4: Qwen3.8-Flash-Next 177B at 11.5 tok/s with expert streaming,RTX 5070 12GB,12,AMD Ryzen 5 5600GT,32 GB DDR4-2400,,Qwen3.8 Flash-Next,"UD-IQ3_XXS, long coding prompt (4,892 tokens out)",llama.cpp-based expert streaming (custom),10.15,none,,,,,yes,https://github.com/ggml-org/llama.cpp/discussions/29930,https://getrigbase.com/builds/rtx-5070-12gb-qwen38-flash-next-expert-streaming/
epyc-7532-rtx-pro-6000-rtx-3090-ti-qwen38-flash-next,"EPYC 7532 workstation (RTX Pro 6000 + RTX 3090 Ti, 300 W caps): Qwen3.8-Flash-Next, llama.cpp vs Strata engine","RTX Pro 6000 Blackwell Workstation (96 GB) + RTX 3090 Ti (EVGA FTW3 Ultra, 24 GB); each test used one card",120,"AMD EPYC 7532 (32 cores, Zen 2)","128 GB DDR4-3200, 8 channels","Board power (not wall), 4K prompt: RTX Pro 6000 297.8 W (llama.cpp) / 290.6 W (Strata); RTX 3090 Ti 168.5 W (llama.cpp) / 281.1 W (Strata)",Qwen3.8 Flash-Next,"IQ3_S, all experts on the RTX Pro 6000, no draft head","llama.cpp (3cf03257), RTX Pro 6000",102.889,speculative,,,,,yes,https://research.strata2signal.com/flash-next-on-three-cards/,https://getrigbase.com/builds/epyc-7532-rtx-pro-6000-rtx-3090-ti-qwen38-flash-next/
epyc-7532-rtx-pro-6000-rtx-3090-ti-qwen38-flash-next,"EPYC 7532 workstation (RTX Pro 6000 + RTX 3090 Ti, 300 W caps): Qwen3.8-Flash-Next, llama.cpp vs Strata engine","RTX Pro 6000 Blackwell Workstation (96 GB) + RTX 3090 Ti (EVGA FTW3 Ultra, 24 GB); each test used one card",120,"AMD EPYC 7532 (32 cores, Zen 2)","128 GB DDR4-3200, 8 channels","Board power (not wall), 4K prompt: RTX Pro 6000 297.8 W (llama.cpp) / 290.6 W (Strata); RTX 3090 Ti 168.5 W (llama.cpp) / 281.1 W (Strata)",Qwen3.8 Flash-Next,"IQ3_S + model draft head, all experts on the RTX Pro 6000","Strata engine 0.1.39, RTX Pro 6000",205.263,speculative,,,,,yes,https://research.strata2signal.com/flash-next-on-three-cards/,https://getrigbase.com/builds/epyc-7532-rtx-pro-6000-rtx-3090-ti-qwen38-flash-next/
epyc-7532-rtx-pro-6000-rtx-3090-ti-qwen38-flash-next,"EPYC 7532 workstation (RTX Pro 6000 + RTX 3090 Ti, 300 W caps): Qwen3.8-Flash-Next, llama.cpp vs Strata engine","RTX Pro 6000 Blackwell Workstation (96 GB) + RTX 3090 Ti (EVGA FTW3 Ultra, 24 GB); each test used one card",120,"AMD EPYC 7532 (32 cores, Zen 2)","128 GB DDR4-3200, 8 channels","Board power (not wall), 4K prompt: RTX Pro 6000 297.8 W (llama.cpp) / 290.6 W (Strata); RTX 3090 Ti 168.5 W (llama.cpp) / 281.1 W (Strata)",Qwen3.8 Flash-Next,"IQ3_S, 32K prompt, all on the RTX Pro 6000","llama.cpp (3cf03257), RTX Pro 6000",76.337,none,,,,,yes,https://research.strata2signal.com/flash-next-on-three-cards/,https://getrigbase.com/builds/epyc-7532-rtx-pro-6000-rtx-3090-ti-qwen38-flash-next/
epyc-7532-rtx-pro-6000-rtx-3090-ti-qwen38-flash-next,"EPYC 7532 workstation (RTX Pro 6000 + RTX 3090 Ti, 300 W caps): Qwen3.8-Flash-Next, llama.cpp vs Strata engine","RTX Pro 6000 Blackwell Workstation (96 GB) + RTX 3090 Ti (EVGA FTW3 Ultra, 24 GB); each test used one card",120,"AMD EPYC 7532 (32 cores, Zen 2)","128 GB DDR4-3200, 8 channels","Board power (not wall), 4K prompt: RTX Pro 6000 297.8 W (llama.cpp) / 290.6 W (Strata); RTX 3090 Ti 168.5 W (llama.cpp) / 281.1 W (Strata)",Qwen3.8 Flash-Next,"IQ3_S, experts of 34/48 layers in system RAM, no draft head","llama.cpp (b11514), RTX 3090 Ti",22.657,speculative,,,,,yes,https://research.strata2signal.com/flash-next-on-three-cards/,https://getrigbase.com/builds/epyc-7532-rtx-pro-6000-rtx-3090-ti-qwen38-flash-next/
epyc-7532-rtx-pro-6000-rtx-3090-ti-qwen38-flash-next,"EPYC 7532 workstation (RTX Pro 6000 + RTX 3090 Ti, 300 W caps): Qwen3.8-Flash-Next, llama.cpp vs Strata engine","RTX Pro 6000 Blackwell Workstation (96 GB) + RTX 3090 Ti (EVGA FTW3 Ultra, 24 GB); each test used one card",120,"AMD EPYC 7532 (32 cores, Zen 2)","128 GB DDR4-3200, 8 channels","Board power (not wall), 4K prompt: RTX Pro 6000 297.8 W (llama.cpp) / 290.6 W (Strata); RTX 3090 Ti 168.5 W (llama.cpp) / 281.1 W (Strata)",Qwen3.8 Flash-Next,"IQ3_S + draft head, 8,144 of 24,576 experts on card","Strata engine 0.1.39, RTX 3090 Ti",98.291,speculative,,,,,yes,https://research.strata2signal.com/flash-next-on-three-cards/,https://getrigbase.com/builds/epyc-7532-rtx-pro-6000-rtx-3090-ti-qwen38-flash-next/
llmspeed-rtx-5090-oct-2026,"RTX 5090 (llm-speed signed runs, Oct 2026): gpt-oss 20B 324 tok/s, Qwen3.8 27B 75 tok/s",RTX 5090,32,AMD EPYC 7543 (Qwen3.8 run: EPYC 9534),,,gpt-oss 20B,reported as Q4_K_M,llama.cpp 0.7.0-dev,324.2,none,,,,,yes,https://llm-speed.com/r/r_dwp5fgdapmw,https://getrigbase.com/builds/llmspeed-rtx-5090-oct-2026/
llmspeed-rtx-5090-oct-2026,"RTX 5090 (llm-speed signed runs, Oct 2026): gpt-oss 20B 324 tok/s, Qwen3.8 27B 75 tok/s",RTX 5090,32,AMD EPYC 7543 (Qwen3.8 run: EPYC 9534),,,Qwen3.6 27B,Q4_K_M,llama.cpp 0.7.0-dev,75.1,none,,,,,yes,https://llm-speed.com/r/r_dwp5fgdapmw,https://getrigbase.com/builds/llmspeed-rtx-5090-oct-2026/
llmspeed-rtx-5090-oct-2026,"RTX 5090 (llm-speed signed runs, Oct 2026): gpt-oss 20B 324 tok/s, Qwen3.8 27B 75 tok/s",RTX 5090,32,AMD EPYC 7543 (Qwen3.8 run: EPYC 9534),,,Qwen3.8 27B,Q4_K_M (EPYC 9534 host),llama.cpp 0.7.0-dev (b11669),75.2,none,,,,,yes,https://llm-speed.com/r/r_dwp5fgdapmw,https://getrigbase.com/builds/llmspeed-rtx-5090-oct-2026/
llmspeed-rtx-4090-oct-2026,"RTX 4090 (llm-speed signed runs, Oct 2026): Qwen3.6 35B-A3B 161 tok/s, Gemma 4 12B QAT 98 tok/s",RTX 4090,24,AMD EPYC 7642 (GLM run: EPYC 75F3),,,Qwen3.6 35B-A3B,Q4_K_M,llama.cpp b8931,160.6,none,,,,,yes,https://llm-speed.com/r/r_m6j1zxwvp-f,https://getrigbase.com/builds/llmspeed-rtx-4090-oct-2026/
llmspeed-rtx-4090-oct-2026,"RTX 4090 (llm-speed signed runs, Oct 2026): Qwen3.6 35B-A3B 161 tok/s, Gemma 4 12B QAT 98 tok/s",RTX 4090,24,AMD EPYC 7642 (GLM run: EPYC 75F3),,,Gemma 4 12B,QAT Q4_0,llama.cpp b8931,97.7,none,,,,,yes,https://llm-speed.com/r/r_m6j1zxwvp-f,https://getrigbase.com/builds/llmspeed-rtx-4090-oct-2026/
llmspeed-rtx-4090-oct-2026,"RTX 4090 (llm-speed signed runs, Oct 2026): Qwen3.6 35B-A3B 161 tok/s, Gemma 4 12B QAT 98 tok/s",RTX 4090,24,AMD EPYC 7642 (GLM run: EPYC 75F3),,,GLM 4.7 Flash,Q4_K_M (EPYC 75F3 host),Ollama 0.31.1,164.9,none,,,,,yes,https://llm-speed.com/r/r_m6j1zxwvp-f,https://getrigbase.com/builds/llmspeed-rtx-4090-oct-2026/
llmspeed-rtx-3090-oct-2026,"RTX 3090 (llm-speed signed runs, Oct 2026): gpt-oss 20B 207 tok/s, Qwen3.6 27B 43 tok/s",RTX 3090,24,AMD EPYC 7282 (GLM run: Threadripper PRO 5975WX),,,gpt-oss 20B,reported as Q4_K_M,llama.cpp 0.7.0-dev,207.2,none,,,,,yes,https://llm-speed.com/r/r_nakw1nq6clw,https://getrigbase.com/builds/llmspeed-rtx-3090-oct-2026/
llmspeed-rtx-3090-oct-2026,"RTX 3090 (llm-speed signed runs, Oct 2026): gpt-oss 20B 207 tok/s, Qwen3.6 27B 43 tok/s",RTX 3090,24,AMD EPYC 7282 (GLM run: Threadripper PRO 5975WX),,,Qwen3.6 27B,Q4_K_M,llama.cpp 0.7.0-dev,42.7,none,,,,,yes,https://llm-speed.com/r/r_nakw1nq6clw,https://getrigbase.com/builds/llmspeed-rtx-3090-oct-2026/
llmspeed-rtx-3090-oct-2026,"RTX 3090 (llm-speed signed runs, Oct 2026): gpt-oss 20B 207 tok/s, Qwen3.6 27B 43 tok/s",RTX 3090,24,AMD EPYC 7282 (GLM run: Threadripper PRO 5975WX),,,GLM 4.7 Flash,Q4_K_M (Threadripper PRO host),Ollama 0.31.1,134.6,none,,,,,yes,https://llm-speed.com/r/r_nakw1nq6clw,https://getrigbase.com/builds/llmspeed-rtx-3090-oct-2026/
llmspeed-2x-rtx-3090-llama33-70b-oct-2026,"2x RTX 3090 (llm-speed signed run, Oct 2026): Llama 3.3 70B Q4_K_M at 19.1 tok/s",2x RTX 3090,48,AMD EPYC 7H12 (64 cores),,,Llama 3.3 70B,Q4_K_M,llama.cpp 0.7.0-dev,19.1,none,,,,,yes,https://llm-speed.com/r/r_n93ukk0mkmm,https://getrigbase.com/builds/llmspeed-2x-rtx-3090-llama33-70b-oct-2026/
minisforum-ms-s1-max-strix-halo-qwen36-35b-vitorpontual,"Minisforum MS-S1 MAX (Strix Halo, 128 GB): Qwen3.6 35B-A3B at 62 tok/s on llama.cpp Vulkan",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,128 GB unified,,Qwen3.6 35B-A3B,Q4_K_M (Unsloth),llama.cpp Vulkan,62,none,,,,,yes,https://vitorpontual.com/posts/two-ai-boxes-3-the-benchmarks/,https://getrigbase.com/builds/minisforum-ms-s1-max-strix-halo-qwen36-35b-vitorpontual/
minisforum-ms-s1-max-strix-halo-qwen36-35b-vitorpontual,"Minisforum MS-S1 MAX (Strix Halo, 128 GB): Qwen3.6 35B-A3B at 62 tok/s on llama.cpp Vulkan",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,128 GB unified,,Qwen3.6 35B-A3B,Q8_0,llama.cpp Vulkan,54,none,,,,,yes,https://vitorpontual.com/posts/two-ai-boxes-3-the-benchmarks/,https://getrigbase.com/builds/minisforum-ms-s1-max-strix-halo-qwen36-35b-vitorpontual/
minisforum-ms-s1-max-strix-halo-qwen36-35b-vitorpontual,"Minisforum MS-S1 MAX (Strix Halo, 128 GB): Qwen3.6 35B-A3B at 62 tok/s on llama.cpp Vulkan",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,128 GB unified,,Qwen3.6 35B-A3B,Q4_K_M (Unsloth),llama.cpp ROCm,51,none,,,,,yes,https://vitorpontual.com/posts/two-ai-boxes-3-the-benchmarks/,https://getrigbase.com/builds/minisforum-ms-s1-max-strix-halo-qwen36-35b-vitorpontual/
minisforum-ms-s1-max-strix-halo-qwen36-35b-vitorpontual,"Minisforum MS-S1 MAX (Strix Halo, 128 GB): Qwen3.6 35B-A3B at 62 tok/s on llama.cpp Vulkan",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,128 GB unified,,Qwen3.6 35B-A3B,Q4_K_M (Unsloth),llama.cpp Vulkan + MTP,93,MTP,,,,,yes,https://vitorpontual.com/posts/two-ai-boxes-3-the-benchmarks/,https://getrigbase.com/builds/minisforum-ms-s1-max-strix-halo-qwen36-35b-vitorpontual/
dgx-spark-qwen36-35b-nvfp4-mtp-vitorpontual,DGX Spark (GB10): Qwen3.6 35B-A3B NVFP4 at 113 tok/s on vLLM with MTP,Nvidia GB10 (unified),128,Nvidia GB10 (20 Arm cores),128 GB unified,,Qwen3.6 35B-A3B,NVFP4,vLLM 0.30 + MTP,113,MTP,,,,,yes,https://vitorpontual.com/posts/two-ai-boxes-3-the-benchmarks/,https://getrigbase.com/builds/dgx-spark-qwen36-35b-nvfp4-mtp-vitorpontual/
dgx-spark-qwen36-35b-nvfp4-mtp-vitorpontual,DGX Spark (GB10): Qwen3.6 35B-A3B NVFP4 at 113 tok/s on vLLM with MTP,Nvidia GB10 (unified),128,Nvidia GB10 (20 Arm cores),128 GB unified,,Qwen3.6 35B-A3B,NVFP4,vLLM 0.30 + MTP,86,MTP,,,,,yes,https://vitorpontual.com/posts/two-ai-boxes-3-the-benchmarks/,https://getrigbase.com/builds/dgx-spark-qwen36-35b-nvfp4-mtp-vitorpontual/
strix-halo-r9700-oculink-strixpro-qwen38-flash-next,Strix Halo + Radeon AI Pro R9700 over OCuLink: Qwen3.8 Flash-Next at ~102 tok/s (custom engine),AMD Radeon AI Pro R9700 32 GB (OCuLink x4) + Ryzen AI Max+ 395 Radeon 8060S iGPU,160,AMD Ryzen AI Max+ (Strix Halo) mini PC,128 GB unified (Strix Halo),R9700 capped at 210 W,Qwen3.8 Flash-Next,custom int4 experts / 6-bit trunk,strixpro (MTP depth 3 + prompt lookup),102,MTP,,,,,yes,https://vuln.dev/flash-engine/,https://getrigbase.com/builds/strix-halo-r9700-oculink-strixpro-qwen38-flash-next/
strix-halo-r9700-oculink-strixpro-qwen38-flash-next,Strix Halo + Radeon AI Pro R9700 over OCuLink: Qwen3.8 Flash-Next at ~102 tok/s (custom engine),AMD Radeon AI Pro R9700 32 GB (OCuLink x4) + Ryzen AI Max+ 395 Radeon 8060S iGPU,160,AMD Ryzen AI Max+ (Strix Halo) mini PC,128 GB unified (Strix Halo),R9700 capped at 210 W,Qwen3.8 Flash-Next,custom int4 experts / 6-bit trunk,"strixpro, ~100K context",82,none,,,,,yes,https://vuln.dev/flash-engine/,https://getrigbase.com/builds/strix-halo-r9700-oculink-strixpro-qwen38-flash-next/
strix-halo-r9700-oculink-strixpro-qwen38-flash-next,Strix Halo + Radeon AI Pro R9700 over OCuLink: Qwen3.8 Flash-Next at ~102 tok/s (custom engine),AMD Radeon AI Pro R9700 32 GB (OCuLink x4) + Ryzen AI Max+ 395 Radeon 8060S iGPU,160,AMD Ryzen AI Max+ (Strix Halo) mini PC,128 GB unified (Strix Halo),R9700 capped at 210 W,Qwen3.8 Flash-Next,custom int4 experts / 6-bit trunk,"strixpro, ~980K context",71,none,,,,,yes,https://vuln.dev/flash-engine/,https://getrigbase.com/builds/strix-halo-r9700-oculink-strixpro-qwen38-flash-next/
strix-halo-r9700-oculink-strixpro-qwen38-flash-next,Strix Halo + Radeon AI Pro R9700 over OCuLink: Qwen3.8 Flash-Next at ~102 tok/s (custom engine),AMD Radeon AI Pro R9700 32 GB (OCuLink x4) + Ryzen AI Max+ 395 Radeon 8060S iGPU,160,AMD Ryzen AI Max+ (Strix Halo) mini PC,128 GB unified (Strix Halo),R9700 capped at 210 W,Qwen3.8 Flash-Next,GGUF,llama.cpp (Strix Halo iGPU only),29-34,none,,,,,yes,https://vuln.dev/flash-engine/,https://getrigbase.com/builds/strix-halo-r9700-oculink-strixpro-qwen38-flash-next/
mac-studio-m5-max-36gb-rapid-mlx-oct-2026,"Mac Studio M5 Max 36 GB: Qwen3.6 35B-A3B at 173 tok/s (Rapid-MLX), 146 on oMLX",Apple M5 Max (unified),36,Apple M5 Max,36 GB unified,,Qwen3.6 35B-A3B,4-bit MLX,Rapid-MLX 0.16.1 (MTP),173.4,MTP,,,,,yes,https://rapidmlx.com/compare,https://getrigbase.com/builds/mac-studio-m5-max-36gb-rapid-mlx-oct-2026/
mac-studio-m5-max-36gb-rapid-mlx-oct-2026,"Mac Studio M5 Max 36 GB: Qwen3.6 35B-A3B at 173 tok/s (Rapid-MLX), 146 on oMLX",Apple M5 Max (unified),36,Apple M5 Max,36 GB unified,,Qwen3.6 35B-A3B,4-bit MLX,oMLX 0.7.0,145.8,none,,,,,yes,https://rapidmlx.com/compare,https://getrigbase.com/builds/mac-studio-m5-max-36gb-rapid-mlx-oct-2026/
mac-studio-m5-max-36gb-rapid-mlx-oct-2026,"Mac Studio M5 Max 36 GB: Qwen3.6 35B-A3B at 173 tok/s (Rapid-MLX), 146 on oMLX",Apple M5 Max (unified),36,Apple M5 Max,36 GB unified,,Qwen3.6 35B-A3B,4-bit MLX,LM Studio (MLX),124.5,none,,,,,yes,https://rapidmlx.com/compare,https://getrigbase.com/builds/mac-studio-m5-max-36gb-rapid-mlx-oct-2026/
mac-studio-m5-max-36gb-rapid-mlx-oct-2026,"Mac Studio M5 Max 36 GB: Qwen3.6 35B-A3B at 173 tok/s (Rapid-MLX), 146 on oMLX",Apple M5 Max (unified),36,Apple M5 Max,36 GB unified,,Qwen3.8 27B,4-bit MLX,Rapid-MLX 0.16.1 (MTP),37.9,MTP,,,,,yes,https://rapidmlx.com/compare,https://getrigbase.com/builds/mac-studio-m5-max-36gb-rapid-mlx-oct-2026/
mac-studio-m5-max-36gb-rapid-mlx-oct-2026,"Mac Studio M5 Max 36 GB: Qwen3.6 35B-A3B at 173 tok/s (Rapid-MLX), 146 on oMLX",Apple M5 Max (unified),36,Apple M5 Max,36 GB unified,,Qwen3.8 27B,4-bit MLX,oMLX 0.7.0,25.3,none,,,,,yes,https://rapidmlx.com/compare,https://getrigbase.com/builds/mac-studio-m5-max-36gb-rapid-mlx-oct-2026/
mac-studio-m5-max-36gb-rapid-mlx-oct-2026,"Mac Studio M5 Max 36 GB: Qwen3.6 35B-A3B at 173 tok/s (Rapid-MLX), 146 on oMLX",Apple M5 Max (unified),36,Apple M5 Max,36 GB unified,,Qwen3.8 27B,4-bit MLX,LM Studio (MLX),26.0,none,,,,,yes,https://rapidmlx.com/compare,https://getrigbase.com/builds/mac-studio-m5-max-36gb-rapid-mlx-oct-2026/
mac-studio-m5-max-36gb-rapid-mlx-oct-2026,"Mac Studio M5 Max 36 GB: Qwen3.6 35B-A3B at 173 tok/s (Rapid-MLX), 146 on oMLX",Apple M5 Max (unified),36,Apple M5 Max,36 GB unified,,Qwen3.5 9B,4-bit MLX,Rapid-MLX 0.16.1 (MTP),115.5,MTP,,,,,yes,https://rapidmlx.com/compare,https://getrigbase.com/builds/mac-studio-m5-max-36gb-rapid-mlx-oct-2026/
mac-studio-m5-max-36gb-rapid-mlx-oct-2026,"Mac Studio M5 Max 36 GB: Qwen3.6 35B-A3B at 173 tok/s (Rapid-MLX), 146 on oMLX",Apple M5 Max (unified),36,Apple M5 Max,36 GB unified,,Qwen3.5 9B,4-bit MLX,oMLX 0.7.0,78.0,none,,,,,yes,https://rapidmlx.com/compare,https://getrigbase.com/builds/mac-studio-m5-max-36gb-rapid-mlx-oct-2026/
macbook-pro-m5-max-128gb-moe-servers-baobab,"MacBook Pro M5 Max 128 GB: Qwen3.6 35B-A3B at 271 tok/s (mlx-serve + MTP), Gemma 4 26B-A4B at 166",Apple M5 Max (unified),128,Apple M5 Max,128 GB unified,,Qwen3.6 35B-A3B,4-bit MLX (mlx-serve build),mlx-serve 26.10.2 + MTP,271,MTP,,,,,yes,https://baobabtech.ai/posts/local-moe-m5-max-qwen-gemma-speed,https://getrigbase.com/builds/macbook-pro-m5-max-128gb-moe-servers-baobab/
macbook-pro-m5-max-128gb-moe-servers-baobab,"MacBook Pro M5 Max 128 GB: Qwen3.6 35B-A3B at 271 tok/s (mlx-serve + MTP), Gemma 4 26B-A4B at 166",Apple M5 Max (unified),128,Apple M5 Max,128 GB unified,,Qwen3.6 35B-A3B,4-bit MLX,Rapid-MLX 0.16.1 + MTP,240,MTP,,,,,yes,https://baobabtech.ai/posts/local-moe-m5-max-qwen-gemma-speed,https://getrigbase.com/builds/macbook-pro-m5-max-128gb-moe-servers-baobab/
macbook-pro-m5-max-128gb-moe-servers-baobab,"MacBook Pro M5 Max 128 GB: Qwen3.6 35B-A3B at 271 tok/s (mlx-serve + MTP), Gemma 4 26B-A4B at 166",Apple M5 Max (unified),128,Apple M5 Max,128 GB unified,,Qwen3.6 35B-A3B,4-bit MLX,mlx_lm.server 0.32.2,107,none,,,,,yes,https://baobabtech.ai/posts/local-moe-m5-max-qwen-gemma-speed,https://getrigbase.com/builds/macbook-pro-m5-max-128gb-moe-servers-baobab/
macbook-pro-m5-max-128gb-moe-servers-baobab,"MacBook Pro M5 Max 128 GB: Qwen3.6 35B-A3B at 271 tok/s (mlx-serve + MTP), Gemma 4 26B-A4B at 166",Apple M5 Max (unified),128,Apple M5 Max,128 GB unified,,Gemma 4 26B-A4B,4-bit MLX (QAT),Rapid-MLX 0.16.1 + DFlash,166,DFlash,,,,,yes,https://baobabtech.ai/posts/local-moe-m5-max-qwen-gemma-speed,https://getrigbase.com/builds/macbook-pro-m5-max-128gb-moe-servers-baobab/
macbook-pro-m5-max-128gb-moe-servers-baobab,"MacBook Pro M5 Max 128 GB: Qwen3.6 35B-A3B at 271 tok/s (mlx-serve + MTP), Gemma 4 26B-A4B at 166",Apple M5 Max (unified),128,Apple M5 Max,128 GB unified,,Qwen3.8 27B,4-bit MLX,"same server, no speculation",31,none,,,,,yes,https://baobabtech.ai/posts/local-moe-m5-max-qwen-gemma-speed,https://getrigbase.com/builds/macbook-pro-m5-max-128gb-moe-servers-baobab/
hp-zbook-ultra-g1a-strix-halo-laptop-petronella,"HP ZBook Ultra G1a (Strix Halo laptop, 128 GB): Qwen3.6 35B-A3B at 59.4 tok/s, gpt-oss 120B at 44.7","AMD Ryzen AI Max+ PRO 395 (Radeon 8060S iGPU, 32 GB dedicated)",128,AMD Ryzen AI Max+ PRO 395,128 GB unified,,Qwen3.6 35B-A3B,Q4_K_M,llama.cpp b11213 Vulkan,59.4,none,,,,,yes,https://petronellatech.com/blog/dgx-spark-vs-strix-halo-vs-gorgon-halo-local-llm-benchmarks/,https://getrigbase.com/builds/hp-zbook-ultra-g1a-strix-halo-laptop-petronella/
hp-zbook-ultra-g1a-strix-halo-laptop-petronella,"HP ZBook Ultra G1a (Strix Halo laptop, 128 GB): Qwen3.6 35B-A3B at 59.4 tok/s, gpt-oss 120B at 44.7","AMD Ryzen AI Max+ PRO 395 (Radeon 8060S iGPU, 32 GB dedicated)",128,AMD Ryzen AI Max+ PRO 395,128 GB unified,,gpt-oss 120B,MXFP4,llama.cpp b11213 Vulkan,44.7,none,,,,,yes,https://petronellatech.com/blog/dgx-spark-vs-strix-halo-vs-gorgon-halo-local-llm-benchmarks/,https://getrigbase.com/builds/hp-zbook-ultra-g1a-strix-halo-laptop-petronella/
hp-zbook-ultra-g1a-strix-halo-laptop-petronella,"HP ZBook Ultra G1a (Strix Halo laptop, 128 GB): Qwen3.6 35B-A3B at 59.4 tok/s, gpt-oss 120B at 44.7","AMD Ryzen AI Max+ PRO 395 (Radeon 8060S iGPU, 32 GB dedicated)",128,AMD Ryzen AI Max+ PRO 395,128 GB unified,,Qwen3.6 35B-A3B,4-bit,Ollama 0.34.0,54.4,none,,,,,yes,https://petronellatech.com/blog/dgx-spark-vs-strix-halo-vs-gorgon-halo-local-llm-benchmarks/,https://getrigbase.com/builds/hp-zbook-ultra-g1a-strix-halo-laptop-petronella/
hp-zbook-ultra-g1a-strix-halo-laptop-petronella,"HP ZBook Ultra G1a (Strix Halo laptop, 128 GB): Qwen3.6 35B-A3B at 59.4 tok/s, gpt-oss 120B at 44.7","AMD Ryzen AI Max+ PRO 395 (Radeon 8060S iGPU, 32 GB dedicated)",128,AMD Ryzen AI Max+ PRO 395,128 GB unified,,Gemma 4 26B-A4B,4-bit,Ollama 0.34.0,51.8,none,,,,,yes,https://petronellatech.com/blog/dgx-spark-vs-strix-halo-vs-gorgon-halo-local-llm-benchmarks/,https://getrigbase.com/builds/hp-zbook-ultra-g1a-strix-halo-laptop-petronella/
hp-zbook-ultra-g1a-strix-halo-laptop-petronella,"HP ZBook Ultra G1a (Strix Halo laptop, 128 GB): Qwen3.6 35B-A3B at 59.4 tok/s, gpt-oss 120B at 44.7","AMD Ryzen AI Max+ PRO 395 (Radeon 8060S iGPU, 32 GB dedicated)",128,AMD Ryzen AI Max+ PRO 395,128 GB unified,,Qwen3.6 27B,4-bit,Ollama 0.34.0,11.4,none,,,,,yes,https://petronellatech.com/blog/dgx-spark-vs-strix-halo-vs-gorgon-halo-local-llm-benchmarks/,https://getrigbase.com/builds/hp-zbook-ultra-g1a-strix-halo-laptop-petronella/
hp-zbook-ultra-g1a-strix-halo-laptop-petronella,"HP ZBook Ultra G1a (Strix Halo laptop, 128 GB): Qwen3.6 35B-A3B at 59.4 tok/s, gpt-oss 120B at 44.7","AMD Ryzen AI Max+ PRO 395 (Radeon 8060S iGPU, 32 GB dedicated)",128,AMD Ryzen AI Max+ PRO 395,128 GB unified,,Gemma 4 31B,4-bit,Ollama 0.34.0,8.8,none,,,,,yes,https://petronellatech.com/blog/dgx-spark-vs-strix-halo-vs-gorgon-halo-local-llm-benchmarks/,https://getrigbase.com/builds/hp-zbook-ultra-g1a-strix-halo-laptop-petronella/
llmspeed-rtx-4090-flint-40b-a2-5b-oct-2026,"RTX 4090 (llm-speed signed run, Oct 2026): Flint 40B-A2.5B at 169-179 tok/s on llama.cpp",RTX 4090,24,AMD EPYC 7452 (32 cores),504 GB,,Flint 40B-A2.5B,NVFP4,llama.cpp 0.7.0-dev (build 11803),179.4,none,,,,,yes,https://llm-speed.com/r/r_sw92j6kg7js,https://getrigbase.com/builds/llmspeed-rtx-4090-flint-40b-a2-5b-oct-2026/
llmspeed-rtx-4090-flint-40b-a2-5b-oct-2026,"RTX 4090 (llm-speed signed run, Oct 2026): Flint 40B-A2.5B at 169-179 tok/s on llama.cpp",RTX 4090,24,AMD EPYC 7452 (32 cores),504 GB,,Flint 40B-A2.5B,UD-Q3_K_XL,llama.cpp 0.7.0-dev (build 11803),169.2,none,,,,,yes,https://llm-speed.com/r/r_sw92j6kg7js,https://getrigbase.com/builds/llmspeed-rtx-4090-flint-40b-a2-5b-oct-2026/
llmspeed-rtx-3090-flint-40b-a2-5b-oct-2026,"RTX 3090 (llm-speed signed run, Oct 2026): Flint 40B-A2.5B Q3_K_XL at 128.9 tok/s",RTX 3090,24,AMD EPYC 7763 (64 cores),1008 GB,,Flint 40B-A2.5B,UD-Q3_K_XL,llama.cpp 0.7.0-dev (build 11803),128.9,none,,,,,yes,https://llm-speed.com/r/r_v-sdiljpm7i,https://getrigbase.com/builds/llmspeed-rtx-3090-flint-40b-a2-5b-oct-2026/
i7-12700k-rtx-5060-ti-rtx-3060-qwen38-27b-jesvs,"i7-12700K + RTX 5060 Ti 16GB + RTX 3060 12GB: Qwen3.8 27B from 21 to 51 tok/s (split, MTP, single-GPU)",RTX 5060 Ti 16GB + RTX 3060 12GB,28,Intel Core i7-12700K,64 GB,,Qwen3.8 27B,"Q4_K_M, layer split 5/3 across both GPUs",llama.cpp (May 2026 build),21.2,none,,,,,yes,https://jesvs.com/posts/llama-cpp-dual-gpu-optimization/,https://getrigbase.com/builds/i7-12700k-rtx-5060-ti-rtx-3060-qwen38-27b-jesvs/
i7-12700k-rtx-5060-ti-rtx-3060-qwen38-27b-jesvs,"i7-12700K + RTX 5060 Ti 16GB + RTX 3060 12GB: Qwen3.8 27B from 21 to 51 tok/s (split, MTP, single-GPU)",RTX 5060 Ti 16GB + RTX 3060 12GB,28,Intel Core i7-12700K,64 GB,,Qwen3.8 27B,"Q4_K_M, split + MTP draft (n_max=2)",llama.cpp master (Oct 2026),37.6,MTP,,,,,yes,https://jesvs.com/posts/llama-cpp-dual-gpu-optimization/,https://getrigbase.com/builds/i7-12700k-rtx-5060-ti-rtx-3060-qwen38-27b-jesvs/
i7-12700k-rtx-5060-ti-rtx-3060-qwen38-27b-jesvs,"i7-12700K + RTX 5060 Ti 16GB + RTX 3060 12GB: Qwen3.8 27B from 21 to 51 tok/s (split, MTP, single-GPU)",RTX 5060 Ti 16GB + RTX 3060 12GB,28,Intel Core i7-12700K,64 GB,,Qwen3.8 27B,"UD-IQ4_XS on the 5060 Ti alone, 8K ctx, q8 KV",llama.cpp master (Oct 2026),28.0,none,,,,,yes,https://jesvs.com/posts/llama-cpp-dual-gpu-optimization/,https://getrigbase.com/builds/i7-12700k-rtx-5060-ti-rtx-3060-qwen38-27b-jesvs/
i7-12700k-rtx-5060-ti-rtx-3060-qwen38-27b-jesvs,"i7-12700K + RTX 5060 Ti 16GB + RTX 3060 12GB: Qwen3.8 27B from 21 to 51 tok/s (split, MTP, single-GPU)",RTX 5060 Ti 16GB + RTX 3060 12GB,28,Intel Core i7-12700K,64 GB,,Qwen3.8 27B,UD-IQ4_XS on the 5060 Ti + MTP draft on the 3060 (n_max=2),llama.cpp master (Oct 2026),51.4,MTP,,,,,yes,https://jesvs.com/posts/llama-cpp-dual-gpu-optimization/,https://getrigbase.com/builds/i7-12700k-rtx-5060-ti-rtx-3060-qwen38-27b-jesvs/
ryzen-9950x3d-rtx-5090-128gb-four-big-moes-gadeyne,Ryzen 9 9950X3D + RTX 5090 + 128 GB DDR5: four big MoEs with experts offloaded to RAM (Qwen3.8 Flash-Next 78.8 tok/s),RTX 5090,32,AMD Ryzen 9 9950X3D (16 cores),128 GB DDR5-5600 (4x32 GB),,Qwen3.8 Flash-Next,"Unsloth UD-Q5_K_XL, MTP drafting, experts offloaded to RAM",Strata (296b59d),78.8,MTP,,,,,yes,https://dev.to/seppegadeyne/four-big-local-llms-on-one-rtx-5090-and-128-gb-of-ram-benchmarked-side-by-side-2h8j,https://getrigbase.com/builds/ryzen-9950x3d-rtx-5090-128gb-four-big-moes-gadeyne/
ryzen-9950x3d-rtx-5090-128gb-four-big-moes-gadeyne,Ryzen 9 9950X3D + RTX 5090 + 128 GB DDR5: four big MoEs with experts offloaded to RAM (Qwen3.8 Flash-Next 78.8 tok/s),RTX 5090,32,AMD Ryzen 9 9950X3D (16 cores),128 GB DDR5-5600 (4x32 GB),,Ornith 1.5 122B,"EXL3 4 bpw, 180/256 experts per layer in RAM, no drafting",TabbyAPI + ExLlamaV3 1.6.2,46.4,none,,,,,yes,https://dev.to/seppegadeyne/four-big-local-llms-on-one-rtx-5090-and-128-gb-of-ram-benchmarked-side-by-side-2h8j,https://getrigbase.com/builds/ryzen-9950x3d-rtx-5090-128gb-four-big-moes-gadeyne/
ryzen-9950x3d-rtx-5090-128gb-four-big-moes-gadeyne,Ryzen 9 9950X3D + RTX 5090 + 128 GB DDR5: four big MoEs with experts offloaded to RAM (Qwen3.8 Flash-Next 78.8 tok/s),RTX 5090,32,AMD Ryzen 9 9950X3D (16 cores),128 GB DDR5-5600 (4x32 GB),,GLM-5.3 Flash,"EXL3 3.05 bpw, MTP drafting, experts offloaded to RAM",Project Maya 1.1.21,30.1,MTP,,,,,yes,https://dev.to/seppegadeyne/four-big-local-llms-on-one-rtx-5090-and-128-gb-of-ram-benchmarked-side-by-side-2h8j,https://getrigbase.com/builds/ryzen-9950x3d-rtx-5090-128gb-four-big-moes-gadeyne/
ryzen-9950x3d-rtx-5090-128gb-four-big-moes-gadeyne,Ryzen 9 9950X3D + RTX 5090 + 128 GB DDR5: four big MoEs with experts offloaded to RAM (Qwen3.8 Flash-Next 78.8 tok/s),RTX 5090,32,AMD Ryzen 9 9950X3D (16 cores),128 GB DDR5-5600 (4x32 GB),,DeepSeek V4.1 Flash,"Maya-S (experts 2-3 bit), partly streamed from SSD",Project Maya 1.1.21 (maya-ds4),24.9,none,,,,,yes,https://dev.to/seppegadeyne/four-big-local-llms-on-one-rtx-5090-and-128-gb-of-ram-benchmarked-side-by-side-2h8j,https://getrigbase.com/builds/ryzen-9950x3d-rtx-5090-128gb-four-big-moes-gadeyne/
rtx-5090-strata-qwen38-flash-next-akitaonrails,RTX 5090 workstation: Qwen3.8 Flash-Next IQ3_S at 159.2 tok/s on Strata (expert offload),RTX 5090,32,,,,Qwen3.8 Flash-Next,"IQ3_S (~55 GB), experts offloaded to RAM, speculation helper",Strata,159.2,none,,,,,yes,https://akitaonrails.github.io/en/2026/10/22/llm-benchmark-v4-consumer-open-source-rtx5090-strix-halo-strata/,https://getrigbase.com/builds/rtx-5090-strata-qwen38-flash-next-akitaonrails/
strix-halo-128gb-qwen38-27b-akitaonrails,"Strix Halo mini PC (Ryzen AI Max+ 395, 128 GB): Qwen3.8 27B at 12.6 tok/s Q4_K_M, 7.7 Q8_0",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,128 GB unified,,Qwen3.8 27B,Q4_K_M,"llama.cpp/Ollama, ROCm",12.6,none,,,,,yes,https://akitaonrails.github.io/en/2026/10/22/llm-benchmark-v4-consumer-open-source-rtx5090-strix-halo-strata/,https://getrigbase.com/builds/strix-halo-128gb-qwen38-27b-akitaonrails/
strix-halo-128gb-qwen38-27b-akitaonrails,"Strix Halo mini PC (Ryzen AI Max+ 395, 128 GB): Qwen3.8 27B at 12.6 tok/s Q4_K_M, 7.7 Q8_0",AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU),128,AMD Ryzen AI Max+ 395,128 GB unified,,Qwen3.8 27B,"Q8_0, --reasoning off","llama.cpp/Ollama, ROCm",7.7,none,,,,,yes,https://akitaonrails.github.io/en/2026/10/22/llm-benchmark-v4-consumer-open-source-rtx5090-strix-halo-strata/,https://getrigbase.com/builds/strix-halo-128gb-qwen38-27b-akitaonrails/
macbook-pro-m4-pro-48gb-flint-40b-kapitsa-rec-llamacpp,"MacBook Pro M4 Pro 48 GB: Flint 40B-A2.5B 4-bit at 127.9 tok/s (Kapitsa), 126.6 (rec), 79.6 (llama.cpp)","Apple M4 Pro (20-core GPU, unified)",48,Apple M4 Pro,48 GB unified,,Flint 40B-A2.5B,int4 MLX (RecursiveSI/flint-int4),Kapitsa (PR #286),127.9,none,,,,,yes,https://github.com/nace-ai/kapitsa/blob/main/bench/results/2026-10-22-flint-m4pro/README.md,https://getrigbase.com/builds/macbook-pro-m4-pro-48gb-flint-40b-kapitsa-rec-llamacpp/
macbook-pro-m4-pro-48gb-flint-40b-kapitsa-rec-llamacpp,"MacBook Pro M4 Pro 48 GB: Flint 40B-A2.5B 4-bit at 127.9 tok/s (Kapitsa), 126.6 (rec), 79.6 (llama.cpp)","Apple M4 Pro (20-core GPU, unified)",48,Apple M4 Pro,48 GB unified,,Flint 40B-A2.5B,int4 MLX (same files),rec 0.1.0 (Recursive),126.56,none,,,,,yes,https://github.com/nace-ai/kapitsa/blob/main/bench/results/2026-10-22-flint-m4pro/README.md,https://getrigbase.com/builds/macbook-pro-m4-pro-48gb-flint-40b-kapitsa-rec-llamacpp/
macbook-pro-m4-pro-48gb-flint-40b-kapitsa-rec-llamacpp,"MacBook Pro M4 Pro 48 GB: Flint 40B-A2.5B 4-bit at 127.9 tok/s (Kapitsa), 126.6 (rec), 79.6 (llama.cpp)","Apple M4 Pro (20-core GPU, unified)",48,Apple M4 Pro,48 GB unified,,Flint 40B-A2.5B,Q4_1 GGUF (RecursiveSI/flint-int4-gguf),"llama.cpp b11804, all layers on GPU, -fa on",79.57,none,,,,,yes,https://github.com/nace-ai/kapitsa/blob/main/bench/results/2026-10-22-flint-m4pro/README.md,https://getrigbase.com/builds/macbook-pro-m4-pro-48gb-flint-40b-kapitsa-rec-llamacpp/
llmspeed-rtx-a4000-muse-glimmer-30b-oct-2026,"RTX A4000 16GB (llm-speed signed run, Oct 2026): Muse Glimmer 30B UD-IQ3_XXS at 21.4 tok/s",RTX A4000,16,,,,Muse Glimmer 30B,UD-IQ3_XXS,llama.cpp 0.7.0-dev (build 11803),21.42,none,,,,,yes,https://llm-speed.com/r/r_y9y09s77lfa,https://getrigbase.com/builds/llmspeed-rtx-a4000-muse-glimmer-30b-oct-2026/
capy-flash-2x-rtx-5070-ti-ryzen-9900x,Dual RTX 5070 Ti on Ryzen 9 9900X (capy-flash maintainer rig),2x RTX 5070 Ti 16 GB,32,AMD Ryzen 9 9900X (12 cores),96 GB DDR5-5600,,Qwen3.8 Flash-Next,GSQ-RCO (64 GB edition),"capy-flash v0.7.0, 2 GPUs, agentic decode",182,none,,,Maintainer/user decode tok/s from the capy-flash README,,yes,https://github.com/ValerioDolci/capy-flash,https://getrigbase.com/builds/capy-flash-2x-rtx-5070-ti-ryzen-9900x/
capy-flash-2x-rtx-5070-ti-ryzen-9900x,Dual RTX 5070 Ti on Ryzen 9 9900X (capy-flash maintainer rig),2x RTX 5070 Ti 16 GB,32,AMD Ryzen 9 9900X (12 cores),96 GB DDR5-5600,,Qwen3.8 Flash-Next,NVFP4,"capy-flash v0.6.0, 2 GPUs, agentic decode",155,none,,,Maintainer/user decode tok/s from the capy-flash README,,yes,https://github.com/ValerioDolci/capy-flash,https://getrigbase.com/builds/capy-flash-2x-rtx-5070-ti-ryzen-9900x/
capy-flash-2x-rtx-5070-ti-ryzen-9900x,Dual RTX 5070 Ti on Ryzen 9 9900X (capy-flash maintainer rig),2x RTX 5070 Ti 16 GB,32,AMD Ryzen 9 9900X (12 cores),96 GB DDR5-5600,,Qwen3.8 27B,3-bit GSQ,"capy-flash (1 GPU), MTP, decode at 8k prompt",139,MTP,,8k prompt,Maintainer/user decode tok/s from the capy-flash README,,yes,https://github.com/ValerioDolci/capy-flash,https://getrigbase.com/builds/capy-flash-2x-rtx-5070-ti-ryzen-9900x/
capy-flash-2x-rtx-5070-ti-ryzen-9900x,Dual RTX 5070 Ti on Ryzen 9 9900X (capy-flash maintainer rig),2x RTX 5070 Ti 16 GB,32,AMD Ryzen 9 9900X (12 cores),96 GB DDR5-5600,,Qwen3.8 27B,3-bit GSQ,"capy-flash, 2 GPUs, MTP, decode at 8k prompt",230,MTP,,8k prompt,Maintainer/user decode tok/s from the capy-flash README,,yes,https://github.com/ValerioDolci/capy-flash,https://getrigbase.com/builds/capy-flash-2x-rtx-5070-ti-ryzen-9900x/
capy-flash-2x-rtx-5080-epyc-7f52,Dual RTX 5080 on EPYC 7F52 (capy-flash user report),2x RTX 5080 16 GB,32,AMD EPYC 7F52 (16-core Zen 2),256 GB DDR4-3200 ECC (8x32 GB),,Qwen3.8 Flash-Next,NVFP4,"capy-flash v0.7.2 (CUDA 13.4), agentic decode",141.3,none,,,"User-reported (issue #11), listed in the capy-flash README",,yes,https://github.com/ValerioDolci/capy-flash/issues/11,https://getrigbase.com/builds/capy-flash-2x-rtx-5080-epyc-7f52/
capy-flash-2x-rtx-5080-epyc-7f52,Dual RTX 5080 on EPYC 7F52 (capy-flash user report),2x RTX 5080 16 GB,32,AMD EPYC 7F52 (16-core Zen 2),256 GB DDR4-3200 ECC (8x32 GB),,Qwen3.8 Flash-Next,NVFP4,"capy-flash v0.7.2, --expert-cpu-threads 15, agentic decode",144.4,none,,,"User-reported (issue #11), listed in the capy-flash README",,yes,https://github.com/ValerioDolci/capy-flash/issues/11,https://getrigbase.com/builds/capy-flash-2x-rtx-5080-epyc-7f52/
capy-flash-2x-rtx-pro-4000-threadripper-5975wx,Dual RTX PRO 4000 Blackwell on Threadripper PRO 5975WX,2x RTX PRO 4000 Blackwell 24 GB,48,AMD Threadripper PRO 5975WX,,145 W per card,Qwen3.8 Flash-Next,GSQ-RCO (64 GB edition),"capy-flash v0.7.1, prose decode (low end of 100-114)",100,none,3300,,User-reported (issue #22); low end of the stated prose range,,yes,https://github.com/ValerioDolci/capy-flash/issues/22,https://getrigbase.com/builds/capy-flash-2x-rtx-pro-4000-threadripper-5975wx/
bnk-2x-tesla-v100-32gb-epyc-7402,Dual Tesla V100 32 GB on EPYC 7402 (bnk engine),"2x Tesla V100 32 GB (PG500-216, sm_70)",64,AMD EPYC 7402 (32 vCPUs),128 GB DDR4 (192 GB for DeepSeek),,Qwen3.8 27B,Q8_0,"bnk dense engine, tensor parallel, no drafting",42.4,none,,,Maintainer decode tok/s from the bnk README,,yes,https://github.com/christ-pher/bnk,https://getrigbase.com/builds/bnk-2x-tesla-v100-32gb-epyc-7402/
bnk-2x-tesla-v100-32gb-epyc-7402,Dual Tesla V100 32 GB on EPYC 7402 (bnk engine),"2x Tesla V100 32 GB (PG500-216, sm_70)",64,AMD EPYC 7402 (32 vCPUs),128 GB DDR4 (192 GB for DeepSeek),,Qwen3.8 27B,Q8_0,"bnk dense engine, tensor parallel, MTP drafting",104,MTP,,,Maintainer decode tok/s from the bnk README,,yes,https://github.com/christ-pher/bnk,https://getrigbase.com/builds/bnk-2x-tesla-v100-32gb-epyc-7402/
bnk-2x-tesla-v100-32gb-epyc-7402,Dual Tesla V100 32 GB on EPYC 7402 (bnk engine),"2x Tesla V100 32 GB (PG500-216, sm_70)",64,AMD EPYC 7402 (32 vCPUs),128 GB DDR4 (192 GB for DeepSeek),,Qwen3.8 Flash-Next,FP4-Local,"bnk v0.3.1, decode at 4K context",114,none,,4K,Maintainer decode tok/s from the bnk README,,yes,https://github.com/christ-pher/bnk,https://getrigbase.com/builds/bnk-2x-tesla-v100-32gb-epyc-7402/
bnk-2x-tesla-v100-32gb-epyc-7402,Dual Tesla V100 32 GB on EPYC 7402 (bnk engine),"2x Tesla V100 32 GB (PG500-216, sm_70)",64,AMD EPYC 7402 (32 vCPUs),128 GB DDR4 (192 GB for DeepSeek),,Qwen3.8 Flash-Next,IQ3_S,"bnk v0.3.1, decode at 4K context",121,none,,4K,Maintainer decode tok/s from the bnk README,,yes,https://github.com/christ-pher/bnk,https://getrigbase.com/builds/bnk-2x-tesla-v100-32gb-epyc-7402/
bnk-2x-tesla-v100-32gb-epyc-7402,Dual Tesla V100 32 GB on EPYC 7402 (bnk engine),"2x Tesla V100 32 GB (PG500-216, sm_70)",64,AMD EPYC 7402 (32 vCPUs),128 GB DDR4 (192 GB for DeepSeek),,DeepSeek V4.1 Flash,GSQ-RCO 2-bit,"bnk v0.5.0, experts split VRAM/RAM (CPU offload)",13,none,,,Maintainer decode tok/s from the bnk README,,yes,https://github.com/christ-pher/bnk,https://getrigbase.com/builds/bnk-2x-tesla-v100-32gb-epyc-7402/
intel-arc-pro-b70-neural-download-lab,Single Intel Arc Pro B70 32 GB (neural.download lab),Intel Arc Pro B70 32 GB,32,,,,Gemma 4 26B-A4B,UD-Q8_K_XL,"llama.cpp SYCL (pinned build, oneAPI 2026.0)",122.16,none,,,"Single-user, cache-cold decode median (lab package/repro README)",,yes,https://github.com/steveseguin/b70-optimization-lab,https://getrigbase.com/builds/intel-arc-pro-b70-neural-download-lab/
intel-arc-pro-b70-neural-download-lab,Single Intel Arc Pro B70 32 GB (neural.download lab),Intel Arc Pro B70 32 GB,32,,,,Ornith 1.5 35B-A3B,Q4_K_M,llama.cpp SYCL + neural.download fusion patches; author marks this headline 'pending',131.46,none,,,"Single-user, cache-cold decode median (lab package/repro README)",pending,yes,https://github.com/steveseguin/b70-optimization-lab,https://getrigbase.com/builds/intel-arc-pro-b70-neural-download-lab/
intel-arc-pro-b70-neural-download-lab,Single Intel Arc Pro B70 32 GB (neural.download lab),Intel Arc Pro B70 32 GB,32,,,,Qwen3.8 27B,Q4_K_M,"llama.cpp SYCL + neural.download TP1 stack, no speculation",27.8,none,,,"Single-user, cache-cold decode median (lab package/repro README)",,yes,https://github.com/steveseguin/b70-optimization-lab,https://getrigbase.com/builds/intel-arc-pro-b70-neural-download-lab/
2x-intel-arc-pro-b70-neural-download-lab,Dual Intel Arc Pro B70 (neural.download lab),2x Intel Arc Pro B70 32 GB,64,,,,Qwen3.8 Flash-Next,IQ3,"neural.download runtime, TP2",58.56,none,,,Lab README/scoreboard,,yes,https://github.com/steveseguin/b70-optimization-lab,https://getrigbase.com/builds/2x-intel-arc-pro-b70-neural-download-lab/
2x-intel-arc-pro-b70-neural-download-lab,Dual Intel Arc Pro B70 (neural.download lab),2x Intel Arc Pro B70 32 GB,64,,,,Qwen3.8 27B,FP8,"vLLM XPU, TP2, MTP 5",46.01,MTP,,,Lab README/scoreboard,,yes,https://github.com/steveseguin/b70-optimization-lab,https://getrigbase.com/builds/2x-intel-arc-pro-b70-neural-download-lab/
intel-arc-pro-b70-vllm-gptq-riceharvest,Single Intel Arc Pro B70 with vLLM XPU cookbook,Intel Arc Pro B70 32 GB,32,,~12 GB free host RAM needed (embeddings in RAM),Stock 275 W cap,Qwen3.8 27B,GPTQ Int4 (asym),"vLLM XPU b70 image + MTP, reasoning prompt",96,MTP,,short prompt,"Greedy, one request at a time, 512-token answers (README ""Numbers"" table)",,yes,https://github.com/riceharvest/b70-qwen38-cookbook,https://getrigbase.com/builds/intel-arc-pro-b70-vllm-gptq-riceharvest/
intel-arc-pro-b70-vllm-gptq-riceharvest,Single Intel Arc Pro B70 with vLLM XPU cookbook,Intel Arc Pro B70 32 GB,32,,~12 GB free host RAM needed (embeddings in RAM),Stock 275 W cap,Qwen3.8 27B,GPTQ Int4 (asym),"vLLM XPU b70 image + MTP, prose prompt",66,MTP,,short prompt,"Greedy, one request at a time, 512-token answers (README ""Numbers"" table)",,yes,https://github.com/riceharvest/b70-qwen38-cookbook,https://getrigbase.com/builds/intel-arc-pro-b70-vllm-gptq-riceharvest/
intel-arc-pro-b70-vllm-gptq-riceharvest,Single Intel Arc Pro B70 with vLLM XPU cookbook,Intel Arc Pro B70 32 GB,32,,~12 GB free host RAM needed (embeddings in RAM),Stock 275 W cap,Qwen3.6 35B-A3B,GPTQ Int4,"vLLM XPU b70 image, no MTP",120,none,,short prompt,"Greedy, one request at a time, 512-token answers (README ""Numbers"" table)",,yes,https://github.com/riceharvest/b70-qwen38-cookbook,https://getrigbase.com/builds/intel-arc-pro-b70-vllm-gptq-riceharvest/
8x-rtx-3090-deepseek-v4-flash-limechain,8x RTX 3090 serving DeepSeek V4 Flash (LimeChain vLLM kit),8x RTX 3090 24 GB,192,,,,DeepSeek V4 Flash,see source (FP8 KV cache),"vLLM LimeChain SM86 overlay, 1 concurrent request",95.52,none,,,"Per-session end-to-end p50 at concurrency 1, 128 generated tokens",,yes,https://github.com/LimeChain/vllm/tree/deepseek-v4-flash-sm86/release/limechain-sm86,https://getrigbase.com/builds/8x-rtx-3090-deepseek-v4-flash-limechain/
4x-dgx-spark-glm-52-quanttrio,4x DGX Spark cluster running GLM-5.2 (unpruned),"4x Nvidia GB10 (unified, DGX Spark)",512,Nvidia Grace (per node),128 GB unified per node,,GLM-5.2,QuantTrio Int4-Int8Mix,"vLLM TP4 + MTP, single-stream mean",32.5,MTP,,200K max,Single-stream mean (source config of 2026-07-17),,yes,https://github.com/tonyd2wild/GLM-5.2-QuantTrio-200K-4x-DGX-Spark,https://getrigbase.com/builds/4x-dgx-spark-glm-52-quanttrio/
4x-dgx-spark-glm-53-flash-ringside,4x DGX Spark ring running GLM-5.3 Flash NVFP4,"4x Nvidia GB10 (unified, DGX Spark)",512,Nvidia Grace 20-core (per node),128 GB unified per node,,GLM-5.3 Flash,NVFP4,"vLLM TP4 (RiNGSiDE), single-stream prose",61.7,none,,,Single-stream decode reported by the source,,yes,https://github.com/othexmr/GLM-5.3-Flash-NVFP4-2x-4x-DGX-Sparks-RiNGSiDE,https://getrigbase.com/builds/4x-dgx-spark-glm-53-flash-ringside/
4x-dgx-spark-glm-53-flash-ringside,4x DGX Spark ring running GLM-5.3 Flash NVFP4,"4x Nvidia GB10 (unified, DGX Spark)",512,Nvidia Grace 20-core (per node),128 GB unified per node,,GLM-5.3 Flash,NVFP4,"vLLM TP4 (RiNGSiDE), single-stream code",105.3,none,,,Single-stream decode reported by the source,,yes,https://github.com/othexmr/GLM-5.3-Flash-NVFP4-2x-4x-DGX-Sparks-RiNGSiDE,https://getrigbase.com/builds/4x-dgx-spark-glm-53-flash-ringside/
2x-dgx-spark-glm-53-flash-ringside,2x DGX Spark running GLM-5.3 Flash NVFP4,"2x Nvidia GB10 (unified, DGX Spark)",256,Nvidia Grace 20-core (per node),128 GB unified per node,,GLM-5.3 Flash,NVFP4,"vLLM TP2 (RiNGSiDE), single-stream prose",33.0,none,,,Single-stream decode reported by the source,,yes,https://github.com/othexmr/GLM-5.3-Flash-NVFP4-2x-4x-DGX-Sparks-RiNGSiDE,https://getrigbase.com/builds/2x-dgx-spark-glm-53-flash-ringside/
2x-dgx-spark-glm-53-flash-ringside,2x DGX Spark running GLM-5.3 Flash NVFP4,"2x Nvidia GB10 (unified, DGX Spark)",256,Nvidia Grace 20-core (per node),128 GB unified per node,,GLM-5.3 Flash,NVFP4,"vLLM TP2 (RiNGSiDE), single-stream code",56.5,none,,,Single-stream decode reported by the source,,yes,https://github.com/othexmr/GLM-5.3-Flash-NVFP4-2x-4x-DGX-Sparks-RiNGSiDE,https://getrigbase.com/builds/2x-dgx-spark-glm-53-flash-ringside/
dgx-spark-qwen36-27b-nvfp4-dflash,DGX Spark with Qwen3.6 27B NVFP4 + DFlash speculative decoding,Nvidia GB10 (unified),128,Nvidia Grace (GB10),128 GB unified,,Qwen3.6 27B,NVFP4,vLLM + DFlash spec decoding,32.83,DFlash,,,"Median decode, Docker reproduction",,yes,https://github.com/my-other-github-account/qwen36-dflash-spark-repro,https://getrigbase.com/builds/dgx-spark-qwen36-27b-nvfp4-dflash/
dgx-spark-qwen36-27b-nvfp4-dflash,DGX Spark with Qwen3.6 27B NVFP4 + DFlash speculative decoding,Nvidia GB10 (unified),128,Nvidia Grace (GB10),128 GB unified,,Qwen3.6 27B,FP8,vLLM autoregressive baseline,7.85,none,,,"Autoregressive baseline, same reproduction",,yes,https://github.com/my-other-github-account/qwen36-dflash-spark-repro,https://getrigbase.com/builds/dgx-spark-qwen36-27b-nvfp4-dflash/
llmspeed-rtx-3090-nemotron-35-bolt,RTX 3090 running Nemotron 3.5 Bolt 74B-A9B in 24 GB (IQ2),RTX 3090,24,,,,Nemotron 3.5 Bolt 74B-A9B,NVIDIA mixed 2-bit IQ2 GGUF (22.0 GB),llama.cpp build 11669 (620d2225),87.75,none,1183.4,4K,"Measured decode, llm-speed run",,yes,https://llm-speed.com/blog/nemotron-3-5-bolt-74b-24gb-measured,https://getrigbase.com/builds/llmspeed-rtx-3090-nemotron-35-bolt/
llmspeed-rtx-4090-bolt-muse-glimmer,RTX 4090: Nemotron 3.5 Bolt (IQ2) and Muse Glimmer 30B,RTX 4090,24,,,,Nemotron 3.5 Bolt 74B-A9B,NVIDIA mixed 2-bit IQ2 GGUF (22.0 GB),llama.cpp build 11669,107.3,none,1814.7,,"Measured decode, llm-speed run",,yes,https://llm-speed.com/blog/nemotron-3-5-bolt-74b-24gb-measured,https://getrigbase.com/builds/llmspeed-rtx-4090-bolt-muse-glimmer/
llmspeed-rtx-4090-bolt-muse-glimmer,RTX 4090: Nemotron 3.5 Bolt (IQ2) and Muse Glimmer 30B,RTX 4090,24,,,,Muse Glimmer 30B,UD-Q4_K_XL,llama.cpp a37cc4610,53.7,none,,8K,"llm-speed can-run page, runs on 2026-10-22, peak 15.1 GiB",,yes,https://llm-speed.com/blog/nemotron-3-5-bolt-74b-24gb-measured,https://getrigbase.com/builds/llmspeed-rtx-4090-bolt-muse-glimmer/
llmspeed-rtx-5090-bolt-qwen38-27b,RTX 5090: Nemotron 3.5 Bolt (IQ2) and Qwen3.8 27B,RTX 5090,32,,,,Nemotron 3.5 Bolt 74B-A9B,NVIDIA mixed 2-bit IQ2 GGUF (22.0 GB),llama.cpp build 11669,154.7,none,2623.3,,"Measured decode, llm-speed run",,yes,https://llm-speed.com/blog/nemotron-3-5-bolt-74b-24gb-measured,https://getrigbase.com/builds/llmspeed-rtx-5090-bolt-qwen38-27b/
llmspeed-rtx-5090-bolt-qwen38-27b,RTX 5090: Nemotron 3.5 Bolt (IQ2) and Qwen3.8 27B,RTX 5090,32,,,,Qwen3.8 27B,Q4_K_M,llama.cpp,75.19,none,,,Comparison run in the same llm-speed post,,yes,https://llm-speed.com/blog/nemotron-3-5-bolt-74b-24gb-measured,https://getrigbase.com/builds/llmspeed-rtx-5090-bolt-qwen38-27b/
omlx-m3-ultra-512gb-minimax-m3,Mac Studio M3 Ultra 512 GB running MiniMax M3 (4-bit MLX),"Apple M3 Ultra (80-core GPU, unified)",512,,512 GB unified,,MiniMax M3,4-bit MLX,oMLX v0.4.4rc2,28.9,none,332.6,1K (20.8 tok/s at 4K),"oMLX community benchmark upload, peak memory 226.6 GB",,yes,https://omlx.ai/benchmarks/performance/bebzj6ar,https://getrigbase.com/builds/omlx-m3-ultra-512gb-minimax-m3/
omlx-m3-ultra-512gb-deepseek-v41-flash-mtp,Mac Studio M3 Ultra 512 GB running DeepSeek V4.1 Flash with MTP,"Apple M3 Ultra (80-core GPU, unified)",512,,512 GB unified,,DeepSeek V4.1 Flash,not stated,"oMLX v0.7.1.dev1, MTP enabled",30.8,MTP,541.7,1K (36.0 tok/s at 8K),"oMLX community benchmark upload, peak memory 289.3 GB",,yes,https://omlx.ai/benchmarks/performance/dvp1bn0s,https://getrigbase.com/builds/omlx-m3-ultra-512gb-deepseek-v41-flash-mtp/
omlx-m3-ultra-256gb-minimax-m31-flash,Mac Studio M3 Ultra (60-core GPU) 256 GB running MiniMax M3.1 Flash,"Apple M3 Ultra (60-core GPU, unified)",256,,256 GB unified,,MiniMax M3.1 Flash,oQ3e-mtp (MTP off),oMLX v0.7.2,25.3,none,290.1,1K (18.0 tok/s at 4K),"oMLX community benchmark upload, mtp_enabled false, peak memory 182.1 GB",,yes,https://omlx.ai/benchmarks/performance/m5fqe90r,https://getrigbase.com/builds/omlx-m3-ultra-256gb-minimax-m31-flash/
omlx-m5-max-128gb-qwen38-27b-mtp,Apple M5 Max (40-core GPU) 128 GB running Qwen3.8 27B with MTP drafter,"Apple M5 Max (40-core GPU, unified)",128,,128 GB unified,,Qwen3.8 27B,4-bit MLX + Qwen3.8-27B-MTP-4bit drafter,"oMLX v0.6.0, MTP enabled",45.3,MTP,963.4,1K (45.8 tok/s at 4K),"oMLX community benchmark upload, peak memory 19.1 GB",,yes,https://omlx.ai/benchmarks/performance/0v8crj8k,https://getrigbase.com/builds/omlx-m5-max-128gb-qwen38-27b-mtp/
c3ma-strix-halo-kolibri-1,"Strix Halo (Ryzen AI Max+ 395, 128 GB) running Kolibri-1 Q4_K_M","AMD Ryzen AI Max+ 395 (Radeon 8060S iGPU, gfx1151)",128,AMD Ryzen AI Max+ 395,128 GB unified,,Kolibri-1,Q4_K_M (47.45 GB),"llama.cpp ROCm (Seraphiel102 fork 8798b65), -ngl 99, f16 KV",58.6,none,951,512 (30.5 tok/s at 131K; prefill measured at 32K),"C3MA AI_Benchmarks repository, commit 22c9f2dcf2 (2026-10-19)",,yes,https://git.mannheim.ccc.de/C3MA/AI_Benchmarks/commits/branch/main,https://getrigbase.com/builds/c3ma-strix-halo-kolibri-1/
four-dgx-spark-step-5-preview,4x DGX Spark (GB10) running Step 5 Preview with tensor parallel over RoCE,"4x Nvidia GB10 (unified, DGX Spark)",512,,128 GB unified per node,,Step 5 Preview,"EXL3 (experts 4.01 bpw, 341.4 GB total)","vLLM, TP=4 over RoCE, MTP (2 draft tokens)",25,MTP,1365,8K,"~25–30 tok/s single stream (published range); prefill 1,365–1,565 tok/s warm at 8K; 66.5 tok/s aggregate with 4 streams",,yes,https://github.com/0xSero/Step-5-Preview-Four-Sparks,https://getrigbase.com/builds/four-dgx-spark-step-5-preview/
lla-rtx-3090-gemma-4-31b,RTX 3090 running Gemma 4 31B Q4_K_M (llama-bench),RTX 3090,24,,,,Gemma 4 31B,Q4_K_M,llama.cpp b11658 (llama-bench),38.8,none,1292,,llama-bench tg128 / pp512; 35.2 tok/s at 16K depth,,yes,https://localllmadvisor.com/can-run/gemma-4-31b-on-rtx-3090-24gb.html,https://getrigbase.com/builds/lla-rtx-3090-gemma-4-31b/
lla-rtx-5090-gemma-4-31b,RTX 5090 running Gemma 4 31B Q4_K_M (llama-bench),RTX 5090,32,,,,Gemma 4 31B,Q4_K_M,llama.cpp b11658 (llama-bench),71.3,none,3841,,llama-bench tg128 / pp512; 64.6 tok/s at 16K depth,,yes,https://localllmadvisor.com/can-run/gemma-4-31b-on-rtx-5090-32gb.html,https://getrigbase.com/builds/lla-rtx-5090-gemma-4-31b/
llmbench-rtx-4070-mellum-21,RTX 4070 12 GB running Mellum 2.1 12B-A2.5B Thinking Q4_K_M,RTX 4070 12GB,12,,,,Mellum 2.1 12B-A2.5B,Thinking GGUF Q4_K_M,llama.cpp,177.5,none,1161.2,131K,"llm-bench.io submission (2026-10-13), memory 8.9 GB",,yes,https://llm-bench.io/benchmarks/cmv76txyv004p01m4h40td1d6,https://getrigbase.com/builds/llmbench-rtx-4070-mellum-21/