IFMTBench - Multi-Constraint Translation Quality: leaderboard
Metric: xCOMET-XXL translation quality (%) on the 2,838 multi-constraint items, translation requests into seven languages with instructions paraphrased in all seven, officially recommended decoding with non-thinking mode for open models; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Hy-MT2-30B-A3B | 74.79 |
| 2 | Gemini 3.1 Pro (Preview) | 69.65 |
| 3 | Gemma 4 26B A4B | 69.61 |
| 4 | Gemma 4 31B | 69.5 |
| 5 | Qwen 3.5 27B (Non-reasoning) | 69.29 |
| 6 | Qwen 3.6 35B A3B (Non-reasoning) | 68.91 |
| 7 | Qwen 3.5 35B A3B (Non-reasoning) | 68.48 |
| 8 | Hy-MT2-1.8B | 67.8 |
| 9 | Gemma 4 E4B | 67.77 |
| 10 | Qwen 3.5 9B (Non-reasoning) | 67.62 |
| 11 | Gemma 4 E2B | 66.92 |
| 12 | Qwen 3.5 4B (Non-reasoning) | 66.44 |
| 13 | Qwen 3.5 2B (Non-reasoning) | 61.83 |
| 14 | Qwen 3.5 0.8B (Non-reasoning) | 52.74 |
Interactive version: theaggregate.ai/benchmark?slug=ifmtbench-multi-constraint-translation-quality · How It Works · Data refreshed daily, snapshot 2026-10-07.