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...
25 cells shown · 24 measured · 1 reported · MIN_RUNS_MEASURED = 3 · capture or correct a number ->
| # | model | rig | basis | value | n | ctx | bits |
|---|---|---|---|---|---|---|---|
| 1 | Qwen3.8-27B-W4A16-AutoRoundchat | CMP 170HX 64GB | measured | 255.9 tok/s | 11 | 131,072 | 4-bit |
| 2 | NVIDIA-Nemotron-3.5-Lightning-30B-A3B-W4A16chat | CMP 170HX 64GB | measured | 226.4 tok/s | 38 | 67,592 | 4-bit |
| 3 | Qwen3.6-35B-A3B-FP8chat | CMP 170HX 64GB | measured | 155.7 tok/s | 78 | 131,335 | 8-bit |
| 4 | Ornith-1.5-35B-A3B-W4A16-SYMchat | CMP 170HX 64GB | measured | 142.5 tok/s | 118 | 131,336 | 4-bit |
| 5 | NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUFchat | CMP 170HX 64GB | measured | 138.7 tok/s | 96 | 67,592 | 8-bit |
| 6 | Ornith-1.5-35B-A3B-MTP-FP8chat | CMP 170HX 64GB | measured | 131.1 tok/s | 76 | 131,336 | 8-bit |
| 7 | Qwen3.6-35B-A3B-GGUFchat | CMP 170HX 64GB | measured | 110.3 tok/s | 203 | 131,338 | 8-bit |
| 8 | Ornith-1.5-35B-A3B-GGUFchat | CMP 170HX 64GB | measured | 108.7 tok/s | 55 | 131,338 | 5-bit |
| 9 | Ornith-1.5-35B-A3B-GGUFchat | CMP 170HX 64GB | measured | 104.8 tok/s | 55 | 131,338 | 8-bit |
| 10 | Ornith-1.5-9Bchat | CMP 170HX 64GB | measured | 72.8 tok/s | 70 | 131,336 | 16-bit |
| 11 | Qwen3.8-27B-W4A16-AutoRound-fastchat | CMP 170HX 64GB | measured | 64 tok/s | 54 | 131,336 | 4-bit |
| 12 | granite-4.2-30b-GGUFchat | CMP 170HX 64GB | measured | 56.2 tok/s | 5 | 65,800 | 16-bit |
| 13 | Qwen3.8-27B-GGUFchat | CMP 170HX 64GB | measured | 49.7 tok/s | 174 | 132,352 | 8-bit |
| 14 | Qwen3.8-27B-GGUFchat | CMP 170HX 64GB | measured | 47 tok/s | 157 | 132,352 | 4-bit |
| 15 | Llama-4-Scout-17B-16E-Instruct-GGUFchat | CMP 170HX 64GB | reported | 44.9 tok/s | 2 | 2,048 | 3-bit |
| 16 | Qwen3.8-Flash-Next-GGUFchat | CMP 170HX 64GB | measured | 44.6 tok/s | 5 | 65,536 | 1-bit |
| 17 | Qwen3.8-27B-MTP-GGUFchat | CMP 170HX 64GB | measured | 42 tok/s | 104 | 131,296 | 4-bit |
| 18 | Qwen3.8-27B-INT8-W8A16-MTPchat | CMP 170HX 64GB | measured | 40.5 tok/s | 127 | 258,311 | 8-bit |
| 19 | Gemma-4-31B-IT-NVFP4chat | CMP 170HX 64GB | measured | 32.9 tok/s | 4 | 32,768 | 4-bit |
| 20 | Qwen3.8-27B-MTP-GGUFchat | CMP 170HX 64GB | measured | 32.7 tok/s | 131 | 132,352 | 8-bit |
| 21 | granite-4.2-30b-GGUFchat | CMP 170HX 64GB | measured | 31.4 tok/s | 48 | 67,589 | 4-bit |
| 22 | granite-4.2-30b-GGUFchat | CMP 170HX 64GB | measured | 27.5 tok/s | 48 | 67,588 | 8-bit |
| 23 | Qwen3.8-27Bchat | CMP 170HX 64GB | measured | 24.4 tok/s | 5 | 65,799 | 16-bit |
| 24 | Llama-3.3-70B-Instruct-NVFP4chat | CMP 170HX 64GB | measured | 23.7 tok/s | 4 | 32,768 | 4-bit |
| 25 | Qwen3.8-27B-GGUFchat | CMP 170HX 64GB | measured | 23.2 tok/s | 43 | 67,549 | 16-bit |
Rows are community cells (medians), ranked provisionally. A signed run outranks any claim.