HELM SeaHELM - IndoNLI: leaderboard
HELM SeaHELM run of IndoNLI, Indonesian natural-language inference (exact match).
Metric: EM. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2 72B Instruct | 86 | #576 |
| 2 | Claude 3.7 Sonnet (20250219) | 85.8 | #201 |
| 3 | Qwen 2.5 72B Instruct | 83.2 | #454 |
| 4 | GPT-4o (2024-11-20) | 83 | #381 |
| 5 | GPT-4o Mini (2024-07-18) | 81.3 | #558 |
| 6 | Gemini 2.0 Flash | 80 | #342 |
| 7 | Gemini 2.0 Flash Lite | 76.8 | #450 |
| 8 | Llama 3.1 405B Instruct | 70.9 | #457 |
| 9 | Llama 3.3 70B Instruct | 52.2 | #534 |
| 10 | Llama 3.1 8B CPT SEA LIONv3 Instruct | 44.6 | #810 |
| 11 | Llama 3.1 70B Instruct | 37.1 | #561 |
| 12 | DeepSeek R1 Distill Llama 8B | 36.5 | #1291 |
| 13 | Llama 4 Scout Instruct | 32.2 | #539 |
| 14 | Gemma 2 9B (IT) | 24.1 | #788 |
| 15 | gemma2-9B-cpt-sea-lionv3-instruct | 15.7 | #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-indonli · How It Works · Data refreshed daily, snapshot 2026-10-20.