The dataset is currently empty. Upload or create new data files. Then, you will be able to explore them in the Dataset Viewer.

Swizzle Runtime (swizzle-rt) Benchmarks

Verified back-to-back inference benchmarks for Swizzle Runtime (v0.1.0), a pure-Rust + CUDA/Vulkan inference engine for GGUF models.

Verified Baseline (AMD Ryzen 7 170 + 4GB RTX 3050 Laptop GPU, 5 reps, bit-identical greedy parity vs llama.cpp)

Model Workload Swizzle Runtime (swizzle-rt) Notes
Qwen3-0.6B-Q8_0 CUDA pp512 11,804.23 tok/s 43.37 ms TTFT
Qwen3-0.6B-Q8_0 CUDA tg128 209.70 tok/s +6.5% vs llama.cpp, ITL p50 4.84 ms
Qwen3-0.6B-Q8_0 CPU edit256 90.61 tok/s +37.7% via RAD-1 self-speculative decoding
LFM3-1.7B-Q4_K_M CUDA tg128 144.26 tok/s +7.0% vs llama.cpp

Quick Install (Linux x86_64, CUDA 12.8)

curl -fsSL https://raw.githubusercontent.com/Swizzle-AI/swizzle-runtime/main/install.sh | sh
swizzle-rt bench model.gguf --device cuda --json
Downloads last month
35