Submit a benchmark
⚡ One command: auto-benchmark NEW
Detects your OS, GPU(s)/Mac/Strix Halo, CPU and RAM, runs llama-bench (pp512 + tg128, -ngl 99), shows you exactly what it would send, and only submits if you type y. No email, no telemetry. Needs Python 3 and llama.cpp.
curl -fsSL https://willitrunllm.com/bench.py | python3 - --model /path/to/model.ggufOther options: --hf-repo bartowski/Qwen_Qwen3-8B-GGUF --hf-file Qwen_Qwen3-8B-Q4_K_M.gguf (downloads via huggingface_hub), --ollama llama3.1:8b (Ollama eval rate, lower fidelity), --dry-run (never sends), --hw <id> to set hardware manually. Read the script first (it's one readable file) or download it and run python3 bench.py --help.
Or paste llama-bench output
Paste the raw markdown table that llama-bench prints, pick the GPU or Mac you ran it on, and check the preview. Submissions are stored privately and only appear on the site after a manual review.
Example: llama-bench -m model-Q4_K_M.gguf -p 512 -n 128. Use one GPU (multi-GPU results aren't shown yet).