Leaderboard
Best local LLM for Thai
Measured Thai scores of quantized local models, per GPU. Highest average so far: gemma4-26b-a4b-qat on AMD Radeon RX 9070 XT.
| Model | GPU | Decode tok/s | Knowledge (translated) % | Local exams % | Math % | Translation chrF++ | Source |
|---|---|---|---|---|---|---|---|
gemma4-26b-a4b-qat | AMD Radeon RX 9070 XT | 90 | - | 65.7 | 89.6 | 55.6 | maintainers |
qwen3.8-27b-q3 | AMD Radeon RX 9070 XT | 50 | - | 67.1 | 87.2 | 54.1 | maintainers |
qwen3.5-9b-q4 | AMD Radeon RX 9070 XT | 42 | - | 61.1 | 83.2 | 52.1 | maintainers |
gemma4-e4b-q4 | AMD Radeon RX 9070 XT | 105 | - | 52.9 | 87.6 | 54.8 | maintainers |
qwen3.5-4b-q4 | AMD Radeon RX 9070 XT | 78 | - | 51.2 | 81.2 | 51.1 | maintainers |
qwen3.8-27b-iq2 | AMD Radeon RX 9070 XT | 40 | - | 54.2 | - | - | maintainers |
Zero-shot, thinking off; margins about ±4-5 points. How each
column is measured: leaderboard. Missing your GPU? Run
localllm eval --langs en,th and share it with localllm report.