HELM SeaHELM - XQuAD (Thai): leaderboard
HELM SeaHELM Southeast-Asian-languages evaluation: Thai cross-lingual extractive question answering.
Metric: SQuAD macro-averaged F1 score. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 4 Maverick Instruct FP8 | 77.14 | #414 |
| 2 | Llama 3.1 8B CPT SEA LIONv3 Instruct | 73.65 | #810 |
| 3 | Llama 3.2 90B Vision Instruct | 72.81 | #622 |
| 4 | Llama 4 Scout Instruct | 71.11 | #539 |
| 5 | Llama 3.1 70B Instruct | 65.23 | #561 |
| 6 | Gemini 2.0 Flash Lite | 64.58 | #450 |
| 7 | Llama 3.3 70B Instruct | 64.44 | #534 |
| 8 | Gemini 2.0 Flash | 63.87 | #342 |
| 9 | Llama 3.1 405B Instruct | 62.41 | #457 |
| 10 | Gemma 2 27B (IT) | 59.03 | #645 |
| 11 | GPT-4o (2024-11-20) | 55.19 | #381 |
| 12 | Claude 3.7 Sonnet (20250219) | 49.78 | #201 |
| 13 | GPT-4o Mini (2024-07-18) | 43.41 | #558 |
| 14 | DeepSeek V3 | 42.46 | #321 |
| 15 | Qwen 2.5 72B Instruct | 38.64 | #454 |
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-xquad-thai · How It Works · Data refreshed daily, snapshot 2026-10-20.