Browse

Community-reported hardware / model / runtime combinations.

Searches GPU, model, and runtime together — try multiple words, e.g. "4090 llama.cpp".

Narrow to one model — combines with search and status. Clear it for all.

2 buckets · one row per Hardware × Model × Runtime combination

Apple M4 Pro × gpt-oss-20b MXFP4 × Ollama

41.0 tok/s

Apple M4 Pro 48GB × gpt-oss-20b MXFP4 × Ollama

42.0 tok/s