HELM SeaHELM - Flores (en-th): leaderboard
HELM SeaHELM Southeast-Asian-languages evaluation: English-to-Thai machine translation.
Metric: ChrF++. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o (2024-11-20) | 49.14 | #381 |
| 2 | Gemini 2.0 Flash | 48.53 | #342 |
| 3 | DeepSeek V3 | 47.01 | #321 |
| 4 | GPT-4o Mini (2024-07-18) | 46.87 | #558 |
| 5 | Llama 4 Maverick Instruct FP8 | 46.1 | #414 |
| 6 | Llama 3.1 405B Instruct | 46.02 | #457 |
| 7 | Claude 3.7 Sonnet (20250219) | 45.97 | #201 |
| 8 | Llama 4 Scout Instruct | 45.09 | #539 |
| 9 | Llama 3.1 70B Instruct | 44.23 | #561 |
| 10 | Llama 3.2 90B Vision Instruct | 44.19 | #622 |
| 11 | Llama 3.3 70B Instruct | 43.95 | #534 |
| 12 | Qwen 2.5 72B Instruct | 43.88 | #454 |
| 13 | Gemini 2.0 Flash Lite | 43.01 | #450 |
| 14 | Qwen 2 72B Instruct | 42.65 | #576 |
| 15 | Claude 3.5 Haiku (20241022) | 41.92 | #588 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-seahelm-flores-en-th · How It Works · Data refreshed daily, snapshot 2026-10-20.