HardMTBench - English to Chinese: leaderboard
Metric: GEMBA-DA direct assessment score (0-100) of the English-to-Chinese translations, judged by gpt-oss-120b, on HardMTBench, 10,000 domain-balanced, difficulty-selected Chinese-English pairs from 12 knowledge-intensive domains, each translated in both directions under one unified protocol; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 92.78 |
| 2 | GPT-5.5 | 92.66 |
| 3 | GPT-5.5 (Non-reasoning) | 92.6 |
| 4 | Qwen 3.5 397B A17B | 92.13 |
| 5 | Qwen 3.5 397B A17B (Non-reasoning) | 91.7 |
| 6 | DeepSeek V4 Pro (Non-reasoning) | 91.33 |
| 7 | Qwen 3.6 35B A3B | 91.05 |
| 8 | Hy-MT2-30B-A3B | 89.71 |
| 9 | Qwen 3.6 35B A3B (Non-reasoning) | 89.13 |
| 10 | Gemma 4 31B (IT) (Thinking) | 89.03 |
| 11 | gemma-4-26B-A4B-it (Thinking) | 88.53 |
| 12 | Gemma 4 31B (IT) | 88.25 |
| 13 | Gemma 4 26B A4B (IT) | 87.57 |
| 14 | Gemma 4 E4B (IT) (Thinking) | 83.24 |
| 15 | gemma-4-E4B-it | 82.82 |
Interactive version: theaggregate.ai/benchmark?slug=hardmtbench-english-to-chinese · How It Works · Data refreshed daily, snapshot 2026-10-07.