Kimi-K3.1-Pro pruned to 179/896 experts, code-calibrated, MXFP4 experts + q4/q6/q8 non-experts. q6 recommended; q4 is fast but lossy.
Pipenetwork PRO
pipenetwork
AI & ML interests
AI & ML
Recent Activity
updated a collection 10 days ago
Kimi-K3.1-Pro REAP80 code (MLX, 512 GB Mac) updated a collection 10 days ago
Kimi-K3.1-Pro REAP80 code (MLX, 512 GB Mac) updated a collection 10 days ago
Kimi-K3.1-Pro REAP80 code (MLX, 512 GB Mac)Organizations
None yet
DeepSeek-V4.1-Flash MLX
Apple Silicon (MLX) builds of DeepSeek-V4.1-Flash (754B) with the only runtime that exists for it: github.com/PipeNetwork/deepseek-v41-mlx
-
pipenetwork/DeepSeek-V4.1-Flash-MLX-mixed-4_8bit
Text Generation • 379B • Updated • 1.17k • 4 -
pipenetwork/DeepSeek-V4.1-Flash-MLX-mixed-4_8bit-engram6
Text Generation • 428B • Updated • 678 • 1 -
pipenetwork/DeepSeek-V4.1-Flash-REAP25-MLX-mixed-4_8bit
Text Generation • 311B • Updated • 477 -
pipenetwork/DeepSeek-V4.1-Flash-REAP50-MLX-mixed-4_8bit
Text Generation • 243B • Updated • 430
GLM-5.3-Flash MLX
Apple Silicon (MLX) builds of GLM-5.3-Flash with a validated, fixed glm5_next runtime: github.com/PipeNetwork/glm53-flash-mlx
-
pipenetwork/GLM-5.3-Flash-MLX-8bit
Image-Text-to-Text • 314B • Updated • 743 • 6 -
pipenetwork/GLM-5.3-Flash-MLX-6bit
Image-Text-to-Text • 314B • Updated • 490 • 4 -
pipenetwork/GLM-5.3-Flash-MLX-mixed-4_8bit
Image-Text-to-Text • 314B • Updated • 2.27k • 15 -
pipenetwork/GLM-5.3-Flash-MLX-4bit
Image-Text-to-Text • 314B • Updated • 736 • 4
Qwen3.8-2.4T-A95B MLX (REAP)
2.4T-param MoE for Apple Silicon. REAP-pruned because no standard quant fits 550GB. Every build measured per-layer against bf16.
MiniMax-H3 MLX
Apple Silicon (MLX) builds of MiniMax-H3, the 33B joint video+audio diffusion transformer. Code: github.com/PipeNetwork/minimax-h3-mlx
-
pipenetwork/MiniMax-H3-MLX-8bit
Image-Text-to-Video • 9B • Updated • 1.17k • 12 -
pipenetwork/MiniMax-H3-MLX-6bit
Image-Text-to-Video • 8B • Updated • 224 • 2 -
pipenetwork/MiniMax-H3-MLX-4bit
Image-Text-to-Video • 7B • Updated • 772 • 7 -
pipenetwork/MiniMax-H3-MLX-bf16
Image-Text-to-Video • 33B • Updated • 821 • 12
Inkling · MLX
975B-A41B multimodal MoE on Apple Silicon. Text+image+audio. Ported from scratch: github.com/PipeNetwork/inkling-mlx
-
pipenetwork/Inkling-MLX-4bit
Image-Text-to-Text • 947B • Updated • 85 • 1 -
pipenetwork/Inkling-MLX-REAP12-4bit
Image-Text-to-Text • 835B • Updated • 42 -
pipenetwork/Inkling-MLX-REAP25-4bit
Image-Text-to-Text • 715B • Updated • 37 -
pipenetwork/Inkling-MLX-REAP50-4bit
Image-Text-to-Text • 483B • Updated • 34
Nemotron TwoTower · MLX
NVIDIA Nemotron TwoTower 30B-A3B diffusion LM in MLX for Apple Silicon: AR tower + two-tower diffusion, 4/6/8-bit + bf16.
-
pipenetwork/Nemotron-3-Nano-30B-A3B-context-mlx-4bit
Text Generation • 32B • Updated • 29 -
pipenetwork/Nemotron-3-Nano-30B-A3B-context-mlx-6bit
Text Generation • 32B • Updated • 35 • 1 -
pipenetwork/Nemotron-3-Nano-30B-A3B-context-mlx-8bit
Text Generation • 32B • Updated • 28 -
pipenetwork/Nemotron-3-Nano-30B-A3B-context-mlx-bf16
Text Generation • 32B • Updated • 28
Ornith-1.0-397B — MLX
MLX (Apple Silicon) conversions of deepreinforce-ai/Ornith-1.0-397B (Qwen3.5-MoE, text-only). 4/6/8-bit and bf16.
GLM-5.2 MLX
First MLX builds of zai-org/GLM-5.2 (glm_moe_dsa MoE): 4/5/6/8-bit + a 512GB-friendly mixed.
Rio-3.1-Open-30B MLX
First MLX (Apple Silicon) quantizations of prefeitura-rio/Rio-3.1-Open-30B (Qwen3-MoE): 4/5/6/8-bit.
Gemma-4-31B-it MLX
MLX (Apple Silicon) quantizations of google/gemma-4-31B-it: 4/5/6/8-bit. Text-only.
MiniMax-M3 MLX
MLX (Apple Silicon) text-only conversions of MiniMax-M3 (427B MoE): 3-bit to 8-bit plus a mixed-precision build.
-
pipenetwork/MiniMax-M3-MLX-8bit
Text Generation • 426B • Updated • 63 • 1 -
pipenetwork/MiniMax-M3-MLX-6bit
Text Generation • 426B • Updated • 47 • 1 -
pipenetwork/MiniMax-M3-MLX-4bit
Text Generation • 426B • Updated • 132 • 1 -
pipenetwork/MiniMax-M3-MLX-mixed-3_6bit
Text Generation • 426B • Updated • 116 • 2
Frog (SWE/debugging) MLX
MLX quants of Microsoft's FrogBoss-32B & FrogMini-14B (Qwen3 debugging finetunes, SWE-bench ~45% pass@1) for Apple Silicon.
-
pipenetwork/FrogMini-14B-2510-MLX-4bit
Text Generation • 15B • Updated • 13 -
pipenetwork/FrogMini-14B-2510-MLX-8bit
Text Generation • 15B • Updated • 13 -
pipenetwork/FrogBoss-32B-2510-MLX-4bit
Text Generation • 33B • Updated • 17 -
pipenetwork/FrogBoss-32B-2510-MLX-8bit
Text Generation • 33B • Updated • 9
Ternary-Bonsai-2 MLX
Stock-runtime MLX builds of Ternary-Bonsai-2-27B: Hadamard rotation unfolded so unmodified mlx-vlm loads it. github.com/PipeNetwork/bonsai2-mlx
-
pipenetwork/Ternary-Bonsai-2-27B-MLX-bf16
Image-Text-to-Text • 27B • Updated • 116 -
pipenetwork/Ternary-Bonsai-2-27B-MLX-8bit
Image-Text-to-Text • 27B • Updated • 129 -
pipenetwork/Ternary-Bonsai-2-27B-MLX-6bit
Image-Text-to-Text • 27B • Updated • 101 -
pipenetwork/Ternary-Bonsai-2-27B-MLX-4bit
Image-Text-to-Text • 27B • Updated • 447 • 2
GLM-5.3 MLX
Apple Silicon (MLX) builds of GLM-5.3 (744B) with a fixed glm_moe_dsa runtime: github.com/PipeNetwork/glm53-mlx
-
pipenetwork/GLM-5.3-MLX-mixed-4_8bit
Text Generation • 743B • Updated • 303 • 2 -
pipenetwork/GLM-5.3-MLX-4bit
Text Generation • 743B • Updated • 339 • 2 -
pipenetwork/GLM-5.3-MLX-mixed-3_6bit
Text Generation • 743B • Updated • 201 • 1 -
pipenetwork/GLM-5.3-MLX-8bit
Text Generation • 743B • Updated • 243 • 1
Qwen3.8-Flash-Next MLX
Apple Silicon (MLX) builds of Qwen3.8-Flash-Next with a validated qwen4_exp runtime: github.com/PipeNetwork/qwen38-flash-next-mlx
-
pipenetwork/Qwen3.8-Flash-Next-MLX-8bit
Text Generation • 177B • Updated • 345 -
pipenetwork/Qwen3.8-Flash-Next-MLX-6bit
Text Generation • 177B • Updated • 295 • 2 -
pipenetwork/Qwen3.8-Flash-Next-MLX-mixed-4_8bit
Text Generation • 182B • Updated • 1.02k • 11 -
pipenetwork/Qwen3.8-Flash-Next-MLX-4bit
Text Generation • 177B • Updated • 330 • 2
Muse-Glimmer 30B MLX
Apple Silicon (MLX) build of Muse-Glimmer-30B, plus the runtime no released mlx-vlm has: github.com/PipeNetwork/muse-glimmer-mlx
Inkling-Small · MLX
276B-A12B multimodal MoE on Apple Silicon. Same code path as Inkling 975B. REAP-25 is free: github.com/PipeNetwork/inkling-mlx
DeepSeek-V4-Flash · MLX
304B MoE on Apple Silicon. deepseek_v4 is in no released runtime — ported from scratch. Take mixed-4_8bit. github.com/PipeNetwork/deepseek-v4-mlx
-
pipenetwork/DeepSeek-V4-Flash-MLX-8bit
Text Generation • 284B • Updated • 68 -
pipenetwork/DeepSeek-V4-Flash-MLX-6bit
Text Generation • 284B • Updated • 51 -
pipenetwork/DeepSeek-V4-Flash-MLX-mixed-4_8bit
Text Generation • 146B • Updated • 142 • 1 -
pipenetwork/DeepSeek-V4-Flash-MLX-4bit
Text Generation • 284B • Updated • 132 • 1
Qwen3.6-35B-A3B — MLX
MLX (Apple Silicon) conversion of Qwen/Qwen3.6-35B-A3B in NVFP4 — the MLX analog of nvidia/Qwen3.6-35B-A3B-NVFP4. Text-only.
GLM-5.2 REAP (expert-pruned)
REAP expert-pruned GLM-5.2 (smaller/faster). REAP-25 near-lossless (+2.3% PPL); REAP-37 +7.3%; REAP-50 smallest but +37.5%.
VISTA MLX
First MLX (Apple Silicon) quantizations of inclusionAI VISTA-9B and VISTA-4B (qwen3_5).
Gemma-4-26B-A4B-it MLX
MLX quantizations of google/gemma-4-26B-A4B-it (MoE): 5/6/8-bit (4-bit available from mlx-community). Text-only.
Kimi-K2.7-Code MLX
MLX build of Kimi-K2.7-Code. Base is natively 4-bit (int4 experts + bf16 rest); this keeps experts at 4-bit and lifts non-expert layers to 6-bit.
Holo-3.1 MLX (computer-use)
First working MLX builds of H Company's Holo-3.1 vision-language computer-use agents (Qwen3.5-VL). Vision-validated. Apache-2.0.
-
pipenetwork/Holo-3.1-4B-MLX-4bit
Image-Text-to-Text • 5B • Updated • 145 • 4 -
pipenetwork/Holo-3.1-4B-MLX-8bit
Image-Text-to-Text • 5B • Updated • 29 • 1 -
pipenetwork/Holo-3.1-9B-MLX-4bit
Image-Text-to-Text • 9B • Updated • 40 -
pipenetwork/Holo-3.1-9B-MLX-8bit
Image-Text-to-Text • 9B • Updated • 30 • 1
Nemotron-3 MLX (Apple Silicon)
MLX quants of NVIDIA Nemotron-3 for Apple Silicon: Ultra 550B (4/5/6/8-bit) and dense Nano-4B (4/8-bit), converted with mlx-lm.
-
pipenetwork/NVIDIA-Nemotron-3-Ultra-550B-A55B-MLX-4bit
Text Generation • 549B • Updated • 51 -
pipenetwork/NVIDIA-Nemotron-3-Ultra-550B-A55B-MLX-8bit
Text Generation • 549B • Updated • 54 -
pipenetwork/NVIDIA-Nemotron-3-Nano-4B-MLX-8bit
1B • Updated • 15 -
pipenetwork/NVIDIA-Nemotron-3-Nano-4B-MLX-4bit
0.6B • Updated • 17
Kimi-K3.1-Pro REAP80 code (MLX, 512 GB Mac)
Kimi-K3.1-Pro pruned to 179/896 experts, code-calibrated, MXFP4 experts + q4/q6/q8 non-experts. q6 recommended; q4 is fast but lossy.
Ternary-Bonsai-2 MLX
Stock-runtime MLX builds of Ternary-Bonsai-2-27B: Hadamard rotation unfolded so unmodified mlx-vlm loads it. github.com/PipeNetwork/bonsai2-mlx
-
pipenetwork/Ternary-Bonsai-2-27B-MLX-bf16
Image-Text-to-Text • 27B • Updated • 116 -
pipenetwork/Ternary-Bonsai-2-27B-MLX-8bit
Image-Text-to-Text • 27B • Updated • 129 -
pipenetwork/Ternary-Bonsai-2-27B-MLX-6bit
Image-Text-to-Text • 27B • Updated • 101 -
pipenetwork/Ternary-Bonsai-2-27B-MLX-4bit
Image-Text-to-Text • 27B • Updated • 447 • 2
DeepSeek-V4.1-Flash MLX
Apple Silicon (MLX) builds of DeepSeek-V4.1-Flash (754B) with the only runtime that exists for it: github.com/PipeNetwork/deepseek-v41-mlx
-
pipenetwork/DeepSeek-V4.1-Flash-MLX-mixed-4_8bit
Text Generation • 379B • Updated • 1.17k • 4 -
pipenetwork/DeepSeek-V4.1-Flash-MLX-mixed-4_8bit-engram6
Text Generation • 428B • Updated • 678 • 1 -
pipenetwork/DeepSeek-V4.1-Flash-REAP25-MLX-mixed-4_8bit
Text Generation • 311B • Updated • 477 -
pipenetwork/DeepSeek-V4.1-Flash-REAP50-MLX-mixed-4_8bit
Text Generation • 243B • Updated • 430
GLM-5.3 MLX
Apple Silicon (MLX) builds of GLM-5.3 (744B) with a fixed glm_moe_dsa runtime: github.com/PipeNetwork/glm53-mlx
-
pipenetwork/GLM-5.3-MLX-mixed-4_8bit
Text Generation • 743B • Updated • 303 • 2 -
pipenetwork/GLM-5.3-MLX-4bit
Text Generation • 743B • Updated • 339 • 2 -
pipenetwork/GLM-5.3-MLX-mixed-3_6bit
Text Generation • 743B • Updated • 201 • 1 -
pipenetwork/GLM-5.3-MLX-8bit
Text Generation • 743B • Updated • 243 • 1
GLM-5.3-Flash MLX
Apple Silicon (MLX) builds of GLM-5.3-Flash with a validated, fixed glm5_next runtime: github.com/PipeNetwork/glm53-flash-mlx
-
pipenetwork/GLM-5.3-Flash-MLX-8bit
Image-Text-to-Text • 314B • Updated • 743 • 6 -
pipenetwork/GLM-5.3-Flash-MLX-6bit
Image-Text-to-Text • 314B • Updated • 490 • 4 -
pipenetwork/GLM-5.3-Flash-MLX-mixed-4_8bit
Image-Text-to-Text • 314B • Updated • 2.27k • 15 -
pipenetwork/GLM-5.3-Flash-MLX-4bit
Image-Text-to-Text • 314B • Updated • 736 • 4
Qwen3.8-Flash-Next MLX
Apple Silicon (MLX) builds of Qwen3.8-Flash-Next with a validated qwen4_exp runtime: github.com/PipeNetwork/qwen38-flash-next-mlx
-
pipenetwork/Qwen3.8-Flash-Next-MLX-8bit
Text Generation • 177B • Updated • 345 -
pipenetwork/Qwen3.8-Flash-Next-MLX-6bit
Text Generation • 177B • Updated • 295 • 2 -
pipenetwork/Qwen3.8-Flash-Next-MLX-mixed-4_8bit
Text Generation • 182B • Updated • 1.02k • 11 -
pipenetwork/Qwen3.8-Flash-Next-MLX-4bit
Text Generation • 177B • Updated • 330 • 2
Qwen3.8-2.4T-A95B MLX (REAP)
2.4T-param MoE for Apple Silicon. REAP-pruned because no standard quant fits 550GB. Every build measured per-layer against bf16.
Muse-Glimmer 30B MLX
Apple Silicon (MLX) build of Muse-Glimmer-30B, plus the runtime no released mlx-vlm has: github.com/PipeNetwork/muse-glimmer-mlx
MiniMax-H3 MLX
Apple Silicon (MLX) builds of MiniMax-H3, the 33B joint video+audio diffusion transformer. Code: github.com/PipeNetwork/minimax-h3-mlx
-
pipenetwork/MiniMax-H3-MLX-8bit
Image-Text-to-Video • 9B • Updated • 1.17k • 12 -
pipenetwork/MiniMax-H3-MLX-6bit
Image-Text-to-Video • 8B • Updated • 224 • 2 -
pipenetwork/MiniMax-H3-MLX-4bit
Image-Text-to-Video • 7B • Updated • 772 • 7 -
pipenetwork/MiniMax-H3-MLX-bf16
Image-Text-to-Video • 33B • Updated • 821 • 12
Inkling-Small · MLX
276B-A12B multimodal MoE on Apple Silicon. Same code path as Inkling 975B. REAP-25 is free: github.com/PipeNetwork/inkling-mlx
Inkling · MLX
975B-A41B multimodal MoE on Apple Silicon. Text+image+audio. Ported from scratch: github.com/PipeNetwork/inkling-mlx
-
pipenetwork/Inkling-MLX-4bit
Image-Text-to-Text • 947B • Updated • 85 • 1 -
pipenetwork/Inkling-MLX-REAP12-4bit
Image-Text-to-Text • 835B • Updated • 42 -
pipenetwork/Inkling-MLX-REAP25-4bit
Image-Text-to-Text • 715B • Updated • 37 -
pipenetwork/Inkling-MLX-REAP50-4bit
Image-Text-to-Text • 483B • Updated • 34
DeepSeek-V4-Flash · MLX
304B MoE on Apple Silicon. deepseek_v4 is in no released runtime — ported from scratch. Take mixed-4_8bit. github.com/PipeNetwork/deepseek-v4-mlx
-
pipenetwork/DeepSeek-V4-Flash-MLX-8bit
Text Generation • 284B • Updated • 68 -
pipenetwork/DeepSeek-V4-Flash-MLX-6bit
Text Generation • 284B • Updated • 51 -
pipenetwork/DeepSeek-V4-Flash-MLX-mixed-4_8bit
Text Generation • 146B • Updated • 142 • 1 -
pipenetwork/DeepSeek-V4-Flash-MLX-4bit
Text Generation • 284B • Updated • 132 • 1
Nemotron TwoTower · MLX
NVIDIA Nemotron TwoTower 30B-A3B diffusion LM in MLX for Apple Silicon: AR tower + two-tower diffusion, 4/6/8-bit + bf16.
-
pipenetwork/Nemotron-3-Nano-30B-A3B-context-mlx-4bit
Text Generation • 32B • Updated • 29 -
pipenetwork/Nemotron-3-Nano-30B-A3B-context-mlx-6bit
Text Generation • 32B • Updated • 35 • 1 -
pipenetwork/Nemotron-3-Nano-30B-A3B-context-mlx-8bit
Text Generation • 32B • Updated • 28 -
pipenetwork/Nemotron-3-Nano-30B-A3B-context-mlx-bf16
Text Generation • 32B • Updated • 28
Qwen3.6-35B-A3B — MLX
MLX (Apple Silicon) conversion of Qwen/Qwen3.6-35B-A3B in NVFP4 — the MLX analog of nvidia/Qwen3.6-35B-A3B-NVFP4. Text-only.
Ornith-1.0-397B — MLX
MLX (Apple Silicon) conversions of deepreinforce-ai/Ornith-1.0-397B (Qwen3.5-MoE, text-only). 4/6/8-bit and bf16.
GLM-5.2 REAP (expert-pruned)
REAP expert-pruned GLM-5.2 (smaller/faster). REAP-25 near-lossless (+2.3% PPL); REAP-37 +7.3%; REAP-50 smallest but +37.5%.
GLM-5.2 MLX
First MLX builds of zai-org/GLM-5.2 (glm_moe_dsa MoE): 4/5/6/8-bit + a 512GB-friendly mixed.
VISTA MLX
First MLX (Apple Silicon) quantizations of inclusionAI VISTA-9B and VISTA-4B (qwen3_5).
Rio-3.1-Open-30B MLX
First MLX (Apple Silicon) quantizations of prefeitura-rio/Rio-3.1-Open-30B (Qwen3-MoE): 4/5/6/8-bit.
Gemma-4-26B-A4B-it MLX
MLX quantizations of google/gemma-4-26B-A4B-it (MoE): 5/6/8-bit (4-bit available from mlx-community). Text-only.
Gemma-4-31B-it MLX
MLX (Apple Silicon) quantizations of google/gemma-4-31B-it: 4/5/6/8-bit. Text-only.
Kimi-K2.7-Code MLX
MLX build of Kimi-K2.7-Code. Base is natively 4-bit (int4 experts + bf16 rest); this keeps experts at 4-bit and lifts non-expert layers to 6-bit.
MiniMax-M3 MLX
MLX (Apple Silicon) text-only conversions of MiniMax-M3 (427B MoE): 3-bit to 8-bit plus a mixed-precision build.
-
pipenetwork/MiniMax-M3-MLX-8bit
Text Generation • 426B • Updated • 63 • 1 -
pipenetwork/MiniMax-M3-MLX-6bit
Text Generation • 426B • Updated • 47 • 1 -
pipenetwork/MiniMax-M3-MLX-4bit
Text Generation • 426B • Updated • 132 • 1 -
pipenetwork/MiniMax-M3-MLX-mixed-3_6bit
Text Generation • 426B • Updated • 116 • 2
Holo-3.1 MLX (computer-use)
First working MLX builds of H Company's Holo-3.1 vision-language computer-use agents (Qwen3.5-VL). Vision-validated. Apache-2.0.
-
pipenetwork/Holo-3.1-4B-MLX-4bit
Image-Text-to-Text • 5B • Updated • 145 • 4 -
pipenetwork/Holo-3.1-4B-MLX-8bit
Image-Text-to-Text • 5B • Updated • 29 • 1 -
pipenetwork/Holo-3.1-9B-MLX-4bit
Image-Text-to-Text • 9B • Updated • 40 -
pipenetwork/Holo-3.1-9B-MLX-8bit
Image-Text-to-Text • 9B • Updated • 30 • 1
Frog (SWE/debugging) MLX
MLX quants of Microsoft's FrogBoss-32B & FrogMini-14B (Qwen3 debugging finetunes, SWE-bench ~45% pass@1) for Apple Silicon.
-
pipenetwork/FrogMini-14B-2510-MLX-4bit
Text Generation • 15B • Updated • 13 -
pipenetwork/FrogMini-14B-2510-MLX-8bit
Text Generation • 15B • Updated • 13 -
pipenetwork/FrogBoss-32B-2510-MLX-4bit
Text Generation • 33B • Updated • 17 -
pipenetwork/FrogBoss-32B-2510-MLX-8bit
Text Generation • 33B • Updated • 9
Nemotron-3 MLX (Apple Silicon)
MLX quants of NVIDIA Nemotron-3 for Apple Silicon: Ultra 550B (4/5/6/8-bit) and dense Nano-4B (4/8-bit), converted with mlx-lm.
-
pipenetwork/NVIDIA-Nemotron-3-Ultra-550B-A55B-MLX-4bit
Text Generation • 549B • Updated • 51 -
pipenetwork/NVIDIA-Nemotron-3-Ultra-550B-A55B-MLX-8bit
Text Generation • 549B • Updated • 54 -
pipenetwork/NVIDIA-Nemotron-3-Nano-4B-MLX-8bit
1B • Updated • 15 -
pipenetwork/NVIDIA-Nemotron-3-Nano-4B-MLX-4bit
0.6B • Updated • 17