HardMTBench - Chinese to English xCOMET: leaderboard
Metric: xCOMET-XXL score (%) of the Chinese-to-English translations, a reference-based learned metric (0-1) times 100, 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: Around 2029. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Hy-MT2-30B-A3B | 69.43 |
| 2 | Qwen 3.5 397B A17B (Non-reasoning) | 65.77 |
| 3 | GPT-5.5 (Non-reasoning) | 65.76 |
| 4 | GPT-5.5 | 65.72 |
| 5 | Gemma 4 31B (IT) (Thinking) | 65.7 |
| 6 | Qwen 3.6 35B A3B | 65.62 |
| 7 | Qwen 3.6 35B A3B (Non-reasoning) | 65.44 |
| 8 | DeepSeek V4 Pro (Non-reasoning) | 65.09 |
| 9 | gemma-4-26B-A4B-it (Thinking) | 64.96 |
| 10 | Gemma 4 31B (IT) | 64.67 |
| 11 | Gemini 3.1 Pro (Preview) | 64.45 |
| 12 | Qwen 3.5 397B A17B | 64.3 |
| 13 | Gemma 4 26B A4B (IT) | 63.81 |
| 14 | Hy-MT2-1.8B | 61.94 |
| 15 | Gemma 4 E4B (IT) (Thinking) | 60.39 |
Interactive version: theaggregate.ai/benchmark?slug=hardmtbench-chinese-to-english-xcomet · How It Works · Data refreshed daily, snapshot 2026-10-07.