BlueBench - Translation: leaderboard
BlueBench translation task using FLORES language pairs to evaluate multilingual translation quality.
Metric: Score (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.3 70B Instruct | 41.71 | #534 |
| 2 | GPT-4o | 41.42 | #343 |
| 3 | GPT-4.1 | 41.22 | #248 |
| 4 | GPT-4.1 Nano | 40.29 | #731 |
| 5 | GPT-4.1 Mini | 40.16 | #355 |
| 6 | Mistral Large | 39.83 | #710 |
| 7 | O3 Mini | 37.79 | #272 |
| 8 | O1 | 37.44 | #183 |
| 9 | O4 Mini | 33.43 | #180 |
| 10 | Pixtral-12B | 32.54 | #808 |
| 11 | Llama 3.2 3B Instruct | 31.01 | #1352 |
| 12 | Llama 3.2 1B Instruct | 23.29 | #1539 |
| 13 | Granite 3.3 8B Instruct | 21.86 | #825 |
| 14 | Mistral Medium 3 | 15.52 | #500 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=bluebench-translation · How It Works · Data refreshed daily, snapshot 2026-10-20.