M4 Max 128 GB

By model

The best run of each model on this machine. Higher is better: fast, and correct.

Best fit among its ranked runs: Qwen3.8-27B-Fast, 49 of 50 tasks at 23.4 tok/s.

1. 2,256
   Qwen3.8-27B-Fastllama.cpp llama-swap (version not exposed) · resyn-bench 2 · 24.8 tok/s · 49/50 tasks · 100k tokens · 2 runs by uwu
2. 44/44
   Qwen3.8-Flash-Nextllama.cpp · resyn.cloud in the browser · 27.6 tok/s · 44/44 tasks · 127k tokens · 1 run by uwu
   Not on the leaderboard: it was posted under an earlier version of the task suite.
3. 42/44
   Qwen3.8-27B-Abliteratedllama.cpp llama-swap (version not exposed) · resyn-bench 2 · 24.4 tok/s · 42/44 tasks · 234k tokens · 1 run by uwu
   Not on the leaderboard: it was posted under an earlier version of the task suite.
4. 41/44
   Qwen3.8-27Bllama.cpp llama-swap (version not exposed) · resyn-bench 2 · 21.9 tok/s · 41/44 tasks · 271k tokens · 1 run by uwu
   Not on the leaderboard: it was posted under an earlier version of the task suite.
5. 41/44
   Qwen3.8-27B-Instructllama.cpp llama-swap (version not exposed) · resyn-bench 2 · 24.5 tok/s · 41/44 tasks · 77k tokens · 1 run by uwu
   Not on the leaderboard: it was posted under an earlier version of the task suite.
6. 29/44
   OpenAI-gpt-oss-120b29/44 tasks · 38k tokens · 1 run by uwu
   Not on the leaderboard: it was posted under an earlier version of the task suite.
7. 27/44
   Gemma-4-31Bllama.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.
8. 27/44
   OpenAI-gpt-oss-20b27/44 tasks · 76k tokens · 1 run by uwu
   Not on the leaderboard: it was posted under an earlier version of the task suite.

Latest runs

1. 41/44
   Qwen3.8-27B-Instruct24.5 tok/s · 77k tokens · llama.cpp · uwu · 21 hours ago
2. 42/44
   Qwen3.8-27B-Fast24.8 tok/s · 58k tokens · llama.cpp · uwu · 22 hours ago
3. 42/44
   Qwen3.8-27B-Abliterated24.4 tok/s · 234k tokens · llama.cpp · uwu · 22 hours ago
4. 41/44
   Qwen3.8-27B21.9 tok/s · 271k tokens · llama.cpp · uwu · 1 day ago
5. 27/44
   OpenAI-gpt-oss-20b111 tok/s · 76k tokens · llama.cpp · uwu · 1 day ago
6. 29/44
   OpenAI-gpt-oss-120b75.9 tok/s · 38k tokens · llama.cpp · uwu · 1 day ago
7. 27/44
   Gemma-4-31B9.8 tok/s · 318k tokens · llama.cpp · uwu · 1 day ago
8. 49/50
   Qwen3.8-27B-Fast23.4 tok/s · 100k tokens · llama.cpp · uwu · 2 days ago
9. 44/44
   Qwen3.8-Flash-Next27.6 tok/s · 127k tokens · llama.cpp · uwu · 5 days ago

Who has one

• uwu · 9 runs