Translation Set1->en COMET22: leaderboard
COMET-22 is a neural machine translation evaluation metric that uses an ensemble of two models: a COMET estimator trained with Direct Assessments and a multitask model that predicts sentence-level scores and word-level OK/BAD tags. It provides improved correlations with human judgments and increased robustness to critical errors compared to previous metrics.
Metric: Set1→en COMET22 (self-reported). Source: benchmarklist.com. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 1.5 Pro (002) | 89.1 |
| 2 | Claude 3.5 Sonnet | 89.1 |
| 3 | GPT-4o | 89 |
| 4 | Nova Pro | 89 |
| 5 | Gemini 1.5 Flash (002) | 88.8 |
| 6 | Nova Lite | 88.8 |
| 7 | GPT-4o Mini | 88.7 |
| 8 | Nova Micro | 88.7 |
| 9 | Gemini 1.5 Flash-8B (001) | 88.5 |
| 10 | Llama 3.2 90B | 88.5 |
| 11 | Claude 3.5 Haiku | 88.3 |
| 12 | Llama 3.1 8B | 86.5 |
Interactive version: theaggregate.ai/benchmark?slug=translation-set1-over-en-comet22 · How It Works · Data refreshed daily, snapshot 2026-09-05.