Qwen3.8-27B-pi
4 records across 2 hardware.
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| A100 PCIe 80 GB | Qwen3.8-27B-pi | IQ2_M | llama.cpp (CUDA) | 48.9 | 945 | – | – | 8192 | 2026-10-06 | community | – | source |
| A100 PCIe 80 GB | Qwen3.8-27B-pi | IQ3_XXS | llama.cpp (CUDA) | 48.8 | 1088 | – | – | 8192 | 2026-10-06 | community | – | source |
| M2 Ultra | Qwen3.8-27B-pi | FP8 | oMLX | 11.8 | 330.3 | – | – | 1024 | 2026-10-07 | community | – | source |
| M2 Ultra | Qwen3.8-27B-pi | FP8 | oMLX | 11.4 | 372.3 | – | – | 4096 | 2026-10-07 | community | – | source |