RX 9070 XT (emulated)
4 records across 1 models (quants: EXL3 3.05 bpw, EXL3 4.05 bpw).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RX 9070 XT (emulated) | Qwen3.8-Flash-Next 80B-A3B | EXL3 3.05 bpw | ranma.cpp (ROCm) | 43.2 | – | – | – | tg128 | 2026-10-05 | community | – | source |
| RX 9070 XT (emulated) | Qwen3.8-Flash-Next 80B-A3B | EXL3 3.05 bpw | ranma.cpp (ROCm) | 39.6 | – | – | – | 65536 | 2026-10-05 | community | – | source |
| RX 9070 XT (emulated) | Qwen3.8-Flash-Next 80B-A3B | EXL3 4.05 bpw | ranma.cpp (ROCm) | 37.0 | – | – | – | tg128 | 2026-10-05 | community | – | source |
| RX 9070 XT (emulated) | Qwen3.8-Flash-Next 80B-A3B | EXL3 4.05 bpw | ranma.cpp (ROCm) | 29.1 | – | – | – | 65536 | 2026-10-05 | community | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085