Free tools
Tools
Standalone tools backed by the llm-speed dataset. No signup, no paywall — every answer cites the runs that back it.
- Will it run — and how fast?/tools/vram-fit
Pick a model, quant, and GPU or Mac — see if it fits in VRAM / unified memory (weights + KV cache), plus the real measured decode tok/s from a signed run for that exact setup.
- Predict tok/s on your rig/tools/predict
Pick a model and a hardware rig — get the predicted decode tokens-per-second from real submitted suite-v1 runs, with the backing runs cited inline.