bestmodel.run / pool — agent view

Honesty ladder: measured > reported > extrapolated > formula > no data yet.
Top 60 cells, default sort: fastest decode. Columns: rig | model | basis | value | n.
Filter: rig=m3-ultra-512gb.
Full machine surface: /llms.txt · REST API: api.bestmodel.run

  rig                               | model                                  | basis    | value          | n
  m3-ultra-512gb                     | qwen-qwen3-1-7b                        | reported | 213 tok/s      | n=1
  m3-ultra-512gb                     | google-gemma-3-4b-it                   | reported | 127.2 tok/s    | n=1
  m3-ultra-512gb                     | nvidia-nemotron-cascade-2-30b-a3b      | reported | 118.1 tok/s    | n=2
  m3-ultra-512gb                     | qwen-qwen3-30b-a3b                     | reported | 102 tok/s      | n=1
  m3-ultra-512gb                     | qwen-qwen3-coder-30b-a3b-instruct      | reported | 100 tok/s      | n=1
  m3-ultra-512gb                     | mlx-community-gemma-4-26b-a4b-it-4bit  | reported | 96.7 tok/s     | n=1
  m3-ultra-512gb                     | ornith-ai-ornith-1-0-35b               | reported | 89.1 tok/s     | n=1
  m3-ultra-512gb                     | qwen-qwen3-8b-gguf                     | reported | 88.3 tok/s     | n=1
  m3-ultra-512gb                     | unsloth-qwen3-6-35b-a3b                | reported | 84.6 tok/s     | n=1
  m3-ultra-512gb                     | ornith-ai-ornith-1-0-9b                | reported | 77.5 tok/s     | n=1
  m3-ultra-512gb                     | google-gemma-4-26b-a4b-it              | reported | 76.9 tok/s     | n=2
  m3-ultra-512gb                     | unsloth-qwen3-coder-next               | reported | 57.1 tok/s     | n=1
  m3-ultra-512gb                     | mistralai-mistral-small-3-2-24b-instruct-2506 | reported | 39.5 tok/s     | n=2
  m3-ultra-512gb                     | mistralai-devstral-small-2505          | reported | 39.4 tok/s     | n=1
  m3-ultra-512gb                     | deepseek-ai-deepseek-v4-flash          | reported | 33 tok/s       | n=1
  m3-ultra-512gb                     | zai-org-glm-5-3-flash                  | reported | 32.9 tok/s     | n=1
  m3-ultra-512gb                     | qwen-qwen3-6-35b-a3b                   | reported | 31.9 tok/s     | n=2
  m3-ultra-512gb                     | unsloth-qwen3-6-27b                    | measured | 26.3 tok/s     | n=3
  m3-ultra-512gb                     | pipenetwork-minimax-m3-mlx-3bit        | reported | 21.2 tok/s     | n=1
  m3-ultra-512gb                     | avlp12-glm-5-2-alis-mlx-dynamic-2-56bpw | reported | 20.1 tok/s     | n=2
  m3-ultra-512gb                     | avlp12-glm-5-2-alis-mlx-dynamic-2-3bpw | reported | 19.8 tok/s     | n=1
  m3-ultra-512gb                     | unsloth-qwen3-6-27b                    | reported | 18.5 tok/s     | n=1
  m3-ultra-512gb                     | avlp12-glm-5-2-alis-mlx-dynamic-3-5bpw | reported | 18.5 tok/s     | n=1
  m3-ultra-512gb                     | qwen-qwen3-235b-a22b-thinking-2507     | measured | 15.6 tok/s     | n=3
  m3-ultra-512gb                     | pipenetwork-glm-5-2-mlx-mixed-3-6bit   | reported | 14.8 tok/s     | n=1
  m3-ultra-512gb                     | qwen-qwen3-8-flash-next                | reported | 14 tok/s       | n=2

Capture or correct a number: /console (signed, Ed25519).