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
12 cells shown · 2 measured · 10 reported · MIN_RUNS_MEASURED = 3 · capture or correct a number ->
| # | model | rig | basis | value | n | ctx | bits |
|---|---|---|---|---|---|---|---|
| 1 | gemma-4-E2B-it-GGUFchat | Ryzen AI Max 385 24GB | measured | 35.2 tok/s | 3 | 2,048 | 4-bit |
| 2 | Qwen3-30B-A3B-GGUFchat | Ryzen AI Max 385 24GB | measured | 30.7 tok/s | 3 | 2,048 | 4-bit |
| 3 | Qwen3-Coder-30B-A3B-Instruct-GGUFcode | Ryzen AI Max 385 24GB | reported | 30 tok/s | 2 | 2,048 | 4-bit |
| 4 | gpt-oss-20bchat | Ryzen AI Max 385 24GB | reported | 23.2 tok/s | 2 | 2,048 | 4-bit |
| 5 | nvidia_Nemotron-Cascade-2-30B-A3B-GGUFchat | Ryzen AI Max 385 24GB | reported | 21.4 tok/s | 2 | 2,048 | 4-bit |
| 6 | Qwen3.6-35B-A3B-GGUFchat | Ryzen AI Max 385 24GB | reported | 19.2 tok/s | 2 | 2,048 | 4-bit |
| 7 | gemma-4-E4B-it-GGUFchat | Ryzen AI Max 385 24GB | reported | 18.7 tok/s | 1 | 2,048 | 4-bit |
| 8 | google_gemma-4-E4B-it-GGUFchat | Ryzen AI Max 385 24GB | reported | 15.4 tok/s | 1 | 2,048 | 4-bit |
| 9 | olmOCR-2-7B-1025-GGUFchat | Ryzen AI Max 385 24GB | reported | 14.9 tok/s | 1 | 2,048 | 4-bit |
| 10 | Qwen3-VL-8B-Instruct-GGUFchat | Ryzen AI Max 385 24GB | reported | 14.1 tok/s | 1 | 2,048 | 4-bit |
| 11 | gemma-4-12b-it-GGUFchat | Ryzen AI Max 385 24GB | reported | 8.6 tok/s | 1 | 2,048 | 4-bit |
| 12 | Qwen3.5-27B-GGUFchat | Ryzen AI Max 385 24GB | reported | 4.1 tok/s | 2 | 2,048 | 4-bit |
Rows are community cells (medians), ranked provisionally. A signed run outranks any claim.