HELM Lite - WMT 2014 (de-en): leaderboard
HELM Lite v1.13.0 WMT 2014 German-to-English machine translation (503 sentences evaluated), 1-shot, scored by BLEU-4 on a 0-100 scale; higher is better. One of the five language pairs averaged into the HELM WMT 2014 board.
Metric: BLEU-4 (%). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 91 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet (20240620) | 22.78 | #308 |
| 2 | Gemini 2.0 Flash (Preview) | 22.58 | #376 |
| 3 | Claude 3.5 Sonnet (20241022) | 22.01 | #302 |
| 4 | Gemini 1.5 Pro (002) | 21.98 | #368 |
| 5 | GPT-4o (2024-05-13) | 21.62 | #337 |
| 6 | Claude 3 Sonnet (20240229) | 21.44 | #694 |
| 7 | Gemini 1.5 Flash (001) | 21.37 | #580 |
| 8 | Claude 3 Opus (20240229) | 21.35 | #428 |
| 9 | Gemini 1.5 Flash (002) | 21.2 | #587 |
| 10 | GPT-4o (2024-08-06) | 20.86 | #343 |
| 11 | Mixtral 8x22B | 20.86 | #725 |
| 12 | Nova Pro | 20.67 | #594 |
| 13 | GPT-4 Turbo | 20.61 | #417 |
| 14 | Llama 3 70B | 20.56 | #632 |
| 15 | Gemma 2 27B (IT) | 20.48 | #652 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-lite-wmt-2014-de-en · How It Works · Data refreshed daily, snapshot 2026-11-05.