ArmBench-LLM - Translation: leaderboard
ArmBench-LLM category: English-Armenian short-text translation scored by BLEU against reference translations. 0-100 BLEU, higher is better.
Metric: BLEU (0-100). Source: huggingface.co. Saturation forecast: Around October 2028. 27 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 31.1 | #90 |
| 2 | Gemma 4 31B (IT) | 28.39 | #178 |
| 3 | Claude Sonnet 4 | 27.82 | #207 |
| 4 | GPT-5.4 Mini | 20.21 | #228 |
| 5 | Gemini 2.5 Flash | 19.9 | #255 |
| 6 | GPT-5.2 Pro | 19.84 | #70 |
| 7 | Grok 4.1 Fast | 19.4 | #227 |
| 8 | Gemini 3 Pro (Preview) | 18.92 | #71 |
| 9 | Gemini 3.1 Pro (Preview) | 18.57 | #64 |
| 10 | Grok 4 Fast | 18.17 | #266 |
| 11 | Claude 3.7 Sonnet | 18.06 | #259 |
| 12 | Gemini 2.5 Pro | 16.69 | #159 |
| 13 | GPT-5.4 Nano | 15.97 | #359 |
| 14 | GLM-5 | 14.11 | #149 |
| 15 | GLM-4.7 Flash | 13.83 | #543 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=armbench-llm-translation · How It Works · Data refreshed daily, snapshot 2026-11-05.