IFMTBench - Multi-Constraint Translation Quality: leaderboard
xCOMET-XXL translation quality (%) on the 2,838 multi-constraint items in IFMTBench (2026), a benchmark of translation requests into seven languages that carry constraints such as schema preservation, glossaries, code spans and register. It measures fidelity to the reference regardless of the constraints; higher is better.
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 September 2028. 15 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Hy-MT2-30B-A3B | 74.79 | ||
| 2 | Hy-MT2-7B | 72.76 | ||
| 3 | Gemini 3.1 Pro (Preview) | 69.65 | #99 | |
| 4 | Gemma 4 26B A4B | 69.61 | #590 | |
| 5 | Gemma 4 31B | 69.5 | #382 | |
| 6 | Qwen 3.5 27B | Non-reasoning | 69.29 | #375 |
| 7 | Qwen 3.6 35B A3B | Non-reasoning | 68.91 | #481 |
| 8 | Qwen 3.5 35B A3B | Non-reasoning | 68.48 | #492 |
| 9 | Hy-MT2-1.8B | 67.8 | ||
| 10 | Gemma 4 E4B | 67.77 | #2184 | |
| 11 | Qwen 3.5 9B | Non-reasoning | 67.62 | #828 |
| 12 | Gemma 4 E2B | 66.92 | #4317 | |
| 13 | Qwen 3.5 4B | Non-reasoning | 66.44 | #1241 |
| 14 | Qwen 3.5 2B | Non-reasoning | 61.83 | #5307 |
| 15 | Qwen 3.5 0.8B | Non-reasoning | 52.74 | #8232 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=ifmtbench-multi-constraint-translation-quality · How It Works · Data refreshed daily, snapshot 2026-11-21.