Leaderboard
Local LLMs on the AMD Radeon RX 9070 XT
Local models measured on the AMD Radeon RX 9070 XT (15.9 GB VRAM): decode speed and the languages each one was scored in.
| Model | Decode tok/s | OS | llama.cpp | Scored in | Source |
|---|---|---|---|---|---|
gemma4-e4b-q4 | 105 | Windows 11 | b11476 | English Arabic Chinese Hindi Japanese Spanish Thai | maintainers |
gemma4-26b-a4b-qat | 90 | Windows 11 | b11476 | English Arabic Chinese Hindi Japanese Spanish Thai | maintainers |
qwen3.5-4b-q4 | 78 | Windows 11 | b11476 | English Arabic Chinese Hindi Japanese Spanish Thai | maintainers |
qwen3.8-27b-q3 | 50 | Windows 11 | b11476 | English Arabic Chinese Hindi Japanese Spanish Thai | maintainers |
qwen3.5-9b-q4 | 42 | Windows 11 | b11476 | English Arabic Chinese Hindi Japanese Spanish Thai | maintainers |
qwen3.8-27b-iq2 | 40 | Windows 11 | b11476 | English Arabic Chinese Hindi Japanese Spanish Thai | maintainers |