MMTIT-Bench - Other to Chinese: leaderboard
Metric: COMET-22 score times 100 (0-100) of end-to-end text-image translation on MMTIT-Bench's 1,400 images (100 each in 14 languages other than English and Chinese; menus, posters, books, documents, products, attractions), against Gemini 2.5 Flash reference translations chosen by model voting and corrected by language experts, identical prompts, final translation only, translating into Chinese; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash | 80.06 | #237 |
| 2 | Qwen 3 VL 235B A22B (Thinking) | 78.81 | #228 (Qwen 3 VL 235B A22B) |
| 3 | Seed-1.6 (Thinking) | 78.77 | #257 (Seed-1.6) |
| 4 | Qwen 3 VL 235B A22B Instruct | 77.2 | #264 |
| 5 | Qwen 2.5 VL 7B Instruct | 71.67 | #643 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=mmtit-bench-other-to-chinese · How It Works · Data refreshed daily, snapshot 2026-10-11.