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

19 cells shown · 1 measured · 18 reported · MIN_RUNS_MEASURED = 3 · capture or correct a number ->

#modelrigbasisvaluenctxbits
1LFM2.5-1.2B-Instruct-GGUFchatM5 Pro 24GBreported267.8 tok/s14,2244-bit
2MiniCPM5-1B-GGUFchatM5 Pro 24GBreported198.2 tok/s14,2248-bit
3LFM2.5-1.2B-Instruct-GGUFchatM5 Pro 24GBreported186 tok/s14,2248-bit
4gemma-4-26B-A4B-it-qat-GGUFchatM5 Pro 24GBreported121.6 tok/s12,0484-bit
5Mellum2-12B-A2.5B-Thinking-GGUF-Q4_K_MchatM5 Pro 24GBreported112.3 tok/s14,2244-bit
6Mellum2-12B-A2.5B-Thinking-GGUF-Q8_0chatM5 Pro 24GBreported86.2 tok/s14,2248-bit
7Qwen3.6-35B-A3B-MTP-GGUFchatM5 Pro 24GBreported80.4 tok/s265,5363-bit
8gemma-4-E2B-it-GGUFchatM5 Pro 24GBreported77.4 tok/s14,2248-bit
9gpt-oss-20b-GGUFchatM5 Pro 24GBreported69.5 tok/s14,2244-bit
10gemma-4-12B-it-qat-GGUFchatM5 Pro 24GBreported67.2 tok/s12,0484-bit
11NVIDIA-Nemotron-3-Nano-4B-GGUFchatM5 Pro 24GBreported63.2 tok/s14,2244-bit
12gemma-4-26B-A4B-it-GGUFchatM5 Pro 24GBreported55 tok/s14,2244-bit
13NVIDIA-Nemotron-3-Nano-4B-GGUFchatM5 Pro 24GBreported53.8 tok/s14,2248-bit
14Qwen3.5-9B-MTP-GGUFchatM5 Pro 24GBreported45.3 tok/s14,2244-bit
15gemma-4-31B-it-qat-GGUFchatM5 Pro 24GBreported27.8 tok/s18,1924-bit
16granite-4.1-8b-GGUFchatM5 Pro 24GBreported26.5 tok/s12,0488-bit
17Qwen3.6-27B-MTP-GGUFchatM5 Pro 24GBreported21.5 tok/s12,0482-bit
18Qwen3.6-27B-MTP-GGUFchatM5 Pro 24GBreported20 tok/s22,0483-bit
19Qwen3.6-27B-MTP-GGUFchatM5 Pro 24GBmeasured18.2 tok/s32,0484-bit

Rows are community cells (medians), ranked provisionally. A signed run outranks any claim.