The pool, cell by cell
pool snapshot 2026-09-18 · every row is a real community cell with its basis and run count · ranking provisional · a signed run outranks any claim
reading the pool...
pool snapshot 2026-09-18 · every row is a real community cell with its basis and run count · ranking provisional · a signed run outranks any claim
reading the pool...
44 cells shown · 24 measured · 20 reported · MIN_RUNS_MEASURED = 3 · capture or correct a number ->
| # | model | rig | basis | value | n | ctx | bits |
|---|---|---|---|---|---|---|---|
| 1 | Qwen3.6-27Bchat | Radeon AI Pro R9700 32GB ×3 | reported | 213.7 tok/s | 1 | 4,096 | 4-bit |
| 2 | gpt-oss-20bchat | Radeon AI Pro R9700 32GB ×3 | measured | 152.9 tok/s | 5 | 131,072 | 8-bit |
| 3 | Laguna-XS-2.1chat | Radeon AI Pro R9700 32GB ×3 | reported | 100.3 tok/s | 1 | 792 | 8-bit |
| 4 | Qwen3.6-35B-A3Bchat | Radeon AI Pro R9700 32GB ×3 | measured | 98.9 tok/s | 3 | 787 | 4-bit |
| 5 | GLM-4.7-Flashchat | Radeon AI Pro R9700 32GB ×3 | reported | 97.4 tok/s | 1 | 783 | 4-bit |
| 6 | Nemotron-Cascade-2-30B-A3Bchat | Radeon AI Pro R9700 32GB ×3 | reported | 89.8 tok/s | 1 | 31,620 | 8-bit |
| 7 | Qwen3.6-35B-A3Bchat | Radeon AI Pro R9700 32GB ×3 | measured | 79.2 tok/s | 5 | 787 | 8-bit |
| 8 | Qwen3-Coder-30B-A3B-Instructcode | Radeon AI Pro R9700 32GB ×3 | reported | 79 tok/s | 2 | 686 | 8-bit |
| 9 | Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressivechat | Radeon AI Pro R9700 32GB ×3 | reported | 76.3 tok/s | 1 | 787 | 8-bit |
| 10 | KAT-Coder-V2.5-Devcode | Radeon AI Pro R9700 32GB ×3 | reported | 75.8 tok/s | 2 | 787 | 8-bit |
| 11 | gemma-4-26B-A4B-itchat | Radeon AI Pro R9700 32GB ×3 | measured | 75.3 tok/s | 4 | 262,144 | 8-bit |
| 12 | Ornith-1.0-35Bchat | Radeon AI Pro R9700 32GB ×3 | reported | 74 tok/s | 2 | 787 | 8-bit |
| 13 | gpt-oss-120bchat | Radeon AI Pro R9700 32GB ×3 | measured | 69.9 tok/s | 4 | 845 | 16-bit |
| 14 | Signal-3.8-27Bchat | Radeon AI Pro R9700 32GB ×3 | measured | 69.4 tok/s | 3 | 262,144 | 8-bit |
| 15 | glm-4-9b-chat-abliteratedchat | Radeon AI Pro R9700 32GB ×3 | reported | 65.8 tok/s | 1 | 783 | 4-bit |
| 16 | Qwen3-Coder-Nextcode | Radeon AI Pro R9700 32GB ×3 | measured | 64.5 tok/s | 4 | 786 | 5-bit |
| 17 | Qwen3-VL-8B-Instructchat | Radeon AI Pro R9700 32GB ×3 | reported | 58.8 tok/s | 1 | 786 | 8-bit |
| 18 | Qwen3.8-27Bchat | Radeon AI Pro R9700 32GB ×3 | measured | 52.9 tok/s | 17 | 262,144 | 8-bit |
| 19 | Muse-Glimmer-30Bchat | Radeon AI Pro R9700 32GB ×3 | reported | 52.8 tok/s | 2 | 1,923 | 4-bit |
| 20 | Laguna-S-2.1chat | Radeon AI Pro R9700 32GB ×3 | measured | 48.3 tok/s | 13 | 131,072 | 4-bit |
| 21 | Ling-3.0-flashchat | Radeon AI Pro R9700 32GB ×3 | measured | 39.3 tok/s | 3 | 4,096 | 4-bit |
| 22 | DeepSeek-V4-Flash-0731chat | Radeon AI Pro R9700 32GB ×3 | measured | 37.3 tok/s | 21 | 32,768 | 1-bit |
| 23 | gemma-4-12B-itchat | Radeon AI Pro R9700 32GB ×3 | reported | 36.7 tok/s | 1 | 794 | 8-bit |
| 24 | Ornith-1.5-35B-A3Bchat | Radeon AI Pro R9700 32GB ×3 | reported | 35.4 tok/s | 1 | 262,144 | 16-bit |
| 25 | MiMo-V2.5chat | Radeon AI Pro R9700 32GB ×3 | measured | 33.2 tok/s | 4 | 32,768 | 2-bit |
| 26 | Qwopus3.5-9B-Codercode | Radeon AI Pro R9700 32GB ×3 | reported | 31.2 tok/s | 1 | 788 | 16-bit |
| 27 | Qwen3.8-Flash-Nextchat | Radeon AI Pro R9700 32GB ×3 | measured | 31.1 tok/s | 3 | 131,072 | 3-bit |
| 28 | MiniMax-M2.5-REAP-139B-A10Bchat | Radeon AI Pro R9700 32GB ×3 | measured | 30.4 tok/s | 5 | 921 | 4-bit |
| 29 | Qwen3.8-27Bchat | Radeon AI Pro R9700 32GB ×3 | measured | 28.4 tok/s | 3 | 829 | 4-bit |
| 30 | Qwen3.5-122B-A10Bchat | Radeon AI Pro R9700 32GB ×3 | measured | 26.5 tok/s | 6 | 65,536 | 4-bit |
| 31 | Ornith-1.0-9Bchat | Radeon AI Pro R9700 32GB ×3 | measured | 25.7 tok/s | 6 | 787 | 16-bit |
| 32 | GLM-5.3-Flashchat | Radeon AI Pro R9700 32GB ×3 | measured | 25.7 tok/s | 5 | 131,072 | 2-bit |
| 33 | MiniMax-M2.7chat | Radeon AI Pro R9700 32GB ×3 | measured | 25.2 tok/s | 3 | 32,768 | 4-bit |
| 34 | Devstral-Small-2-24B-Instruct-2512chat | Radeon AI Pro R9700 32GB ×3 | reported | 23.9 tok/s | 1 | 24,576 | 8-bit |
| 35 | Qwen3-32Bchat | Radeon AI Pro R9700 32GB ×3 | reported | 22.8 tok/s | 1 | 786 | 8-bit |
| 36 | Muse-Glimmer-30Bchat | Radeon AI Pro R9700 32GB ×3 | measured | 21.3 tok/s | 8 | 1,923 | 8-bit |
| 37 | gemma-4-12B-itchat | Radeon AI Pro R9700 32GB ×3 | reported | 19.3 tok/s | 2 | 794 | 16-bit |
| 38 | Qwen3.6-27Bchat | Radeon AI Pro R9700 32GB ×3 | measured | 17.8 tok/s | 19 | 262,144 | 8-bit |
| 39 | granite-4.1-30bchat | Radeon AI Pro R9700 32GB ×3 | reported | 17.3 tok/s | 1 | 65,536 | 8-bit |
| 40 | gemma-4-31B-itchat | Radeon AI Pro R9700 32GB ×3 | measured | 14.2 tok/s | 4 | 794 | 8-bit |
| 41 | Kimi-Dev-72Bchat | Radeon AI Pro R9700 32GB ×3 | reported | 11.4 tok/s | 1 | 65,536 | 4-bit |
| 42 | DeepSeek-V4-Flashchat | Radeon AI Pro R9700 32GB ×3 | measured | 9.2 tok/s | 9 | 8,192 | 1-bit |
| 43 | Qwen3.8-Flash-Nextchat | Radeon AI Pro R9700 32GB ×3 | reported | 7.7 tok/s | 1 | 789 | 4-bit |
| 44 | Mistral-Medium-3.5-128Bchat | Radeon AI Pro R9700 32GB ×3 | measured | 6.4 tok/s | 9 | 16,384 | 4-bit |
Rows are community cells (medians), ranked provisionally. A signed run outranks any claim.