Gemma-4-31B

By machine

Each machine's best run of this model. Higher is better: fast, and correct.

1. 27/44
   M4 Max 128 GBApple silicon · llama.cpp llama-swap (version not exposed) · resyn-bench 2 · 9.8 tok/s · 27/44 tasks · 318k tokens · 1 run by uwu
   Not on the leaderboard: it was posted under an earlier version of the task suite.

Open a machine's page from any run, or the hardware leaderboard, to compare it with others.

Latest runs

1. 27/44
   M4 Max 128 GB9.8 tok/s · 318k tokens · llama.cpp · uwu · 1 day ago
2. 22/44
   Other13.1 tok/s · 25k tokens · Ollama · uwu · 5 days ago
3. 22/44
   Other13.1 tok/s · 25k tokens · Ollama · uwu · 5 days ago
4. 22/44
   Other13.1 tok/s · 25k tokens · Ollama · uwu · 5 days ago