HELM SeaHELM - IndicQA: leaderboard
HELM SeaHELM Southeast-Asian-languages evaluation: Indic-language reading-comprehension 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 3.2 90B Vision Instruct | 57.23 | #622 |
| 2 | Llama 4 Maverick Instruct FP8 | 48.42 | #414 |
| 3 | Gemma 2 27B (IT) | 47.24 | #645 |
| 4 | Llama 3.1 70B Instruct | 43.31 | #561 |
| 5 | Llama 3.3 70B Instruct | 39.05 | #534 |
| 6 | GPT-4o (2024-11-20) | 35.44 | #381 |
| 7 | Llama 4 Scout Instruct | 34 | #539 |
| 8 | Llama 3.1 405B Instruct | 32.7 | #457 |
| 9 | Gemini 2.0 Flash Lite | 31.54 | #450 |
| 10 | Gemini 2.0 Flash | 29.76 | #342 |
| 11 | GPT-4o Mini (2024-07-18) | 29.39 | #558 |
| 12 | Qwen 2 72B Instruct | 27.21 | #576 |
| 13 | Gemma 2 9B (IT) | 26.25 | #788 |
| 14 | DeepSeek V3 | 22.69 | #321 |
| 15 | gemma2-9B-cpt-sea-lionv3-instruct | 21.72 | #705 |
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-indicqa · How It Works · Data refreshed daily, snapshot 2026-10-20.