IFMTBench - Single-Constraint Translation Quality: leaderboard
Metric: xCOMET-XXL translation quality (%) on the 4,506 single-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 January 2027. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Hy-MT2-30B-A3B | 85.55 |
| 2 | Gemma 4 31B | 83.56 |
| 3 | Gemini 3.1 Pro (Preview) | 83.31 |
| 4 | Gemma 4 26B A4B | 83.3 |
| 5 | Qwen 3.5 27B (Non-reasoning) | 82.64 |
| 6 | Qwen 3.6 35B A3B (Non-reasoning) | 82.56 |
| 7 | Gemma 4 E4B | 82.26 |
| 8 | Qwen 3.5 35B A3B (Non-reasoning) | 81.74 |
| 9 | Qwen 3.5 9B (Non-reasoning) | 80.8 |
| 10 | Hy-MT2-1.8B | 80.74 |
| 11 | Gemma 4 E2B | 80.22 |
| 12 | Qwen 3.5 4B (Non-reasoning) | 80.16 |
| 13 | Qwen 3.5 2B (Non-reasoning) | 75.77 |
| 14 | Qwen 3.5 0.8B (Non-reasoning) | 70.21 |
Interactive version: theaggregate.ai/benchmark?slug=ifmtbench-single-constraint-translation-quality · How It Works · Data refreshed daily, snapshot 2026-10-07.