IFMTBench: leaderboard
Modern translation workflows demand more than semantic equivalence. Users routinely require models to preserve JSON or HTML schemas, honor curated glossaries, disambiguate with provided context, and match prescribed registers, often several at once. Conventional metrics such as BLEU and xCOMET capture semantic fidelity but provide little signal on constraint adherence, while general instruction fo
Metric: IF$_\text{T}$ (self-reported). Source: benchmarklist.com. Saturation forecast: Around June 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 89.08 | #58 |
| 2 | Hy-MT2-30B-A3B | 85.7 | #359 |
| 3 | Qwen 3.5 27B | 81.06 | #187 |
| 4 | Gemma 4 31B | 79.61 | #200 |
| 5 | Gemma 4 E4B | 72.56 | #767 |
| 6 | Qwen 3.5 9B | 71.22 | #379 |
| 7 | Qwen 3.5 4B | 66.65 | #475 |
| 8 | Gemma 4 E2B | 58.69 | #1070 |
| 9 | Qwen 3.5 2B | 43.11 | #1148 |
| 10 | Qwen 3.5 0.8B | 27.99 | #1424 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ifmtbench · How It Works · Data refreshed daily, snapshot 2026-10-20.