HardMTBench - English to Chinese xCOMET: leaderboard
Metric: xCOMET-XXL score (%) of the English-to-Chinese 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 February 2028. 22 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Hy-MT2-30B-A3B | 75.87 |
| 2 | Gemini 3.1 Pro (Preview) | 74.15 |
| 3 | GPT-5.5 | 73.44 |
| 4 | GPT-5.5 (Non-reasoning) | 73.33 |
| 5 | Qwen 3.6 35B A3B | 72.57 |
| 6 | Qwen 3.5 397B A17B | 72.36 |
| 7 | Qwen 3.5 397B A17B (Non-reasoning) | 72.26 |
| 8 | Gemma 4 31B (IT) (Thinking) | 72.1 |
| 9 | gemma-4-26B-A4B-it (Thinking) | 71.81 |
| 10 | Qwen 3.6 35B A3B (Non-reasoning) | 71.8 |
| 11 | Gemma 4 26B A4B (IT) | 71.09 |
| 12 | Gemma 4 31B (IT) | 71.04 |
| 13 | DeepSeek V4 Pro (Non-reasoning) | 70.84 |
| 14 | Hy-MT2-1.8B | 70.72 |
| 15 | Gemma 4 E4B (IT) (Thinking) | 67.35 |
Interactive version: theaggregate.ai/benchmark?slug=hardmtbench-english-to-chinese-xcomet · How It Works · Data refreshed daily, snapshot 2026-10-07.