7900 XTX
7 records across 1 models (quants: UD-IQ3_XXS, UD-IQ4_XS).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 7900 XTX | Qwen3.8-Flash-Next 125B | UD-IQ3_XXS | llama.cpp (ROCm) | 44.8 | – | – | – | 4096 | 2026-10-04 | community | – | source |
| 7900 XTX | Qwen3.8-Flash-Next 125B | UD-IQ3_XXS | llama.cpp (ROCm) | 37.2 | – | – | – | 4096 | 2026-10-04 | community | – | source |
| 7900 XTX | Qwen3.8-Flash-Next 125B | UD-IQ4_XS | llama.cpp (ROCm) | 31.2 | – | – | – | 4096 | 2026-10-04 | community | – | source |
| 7900 XTX | Qwen3.8-Flash-Next 125B | UD-IQ4_XS | llama.cpp (ROCm) | 30.5 | – | – | – | 4096 | 2026-10-04 | community | – | source |
| 7900 XTX | Qwen3.8-Flash-Next 125B | UD-IQ4_XS | llama.cpp (ROCm) | 28.8 | – | – | – | 4096 | 2026-10-04 | community | – | source |
| 7900 XTX | Qwen3.8-Flash-Next 125B | UD-IQ3_XXS | llama.cpp (ROCm) | 11.3 | – | – | – | 4096 | 2026-10-04 | community | – | source |
| 7900 XTX | Qwen3.8-Flash-Next 125B | UD-IQ3_XXS | llama.cpp (ROCm) | 9.7 | – | – | – | 4096 | 2026-10-04 | community | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085