HELM SeaHELM - NusaX — leaderboard
Metric: Macro F1 score. Source: crfm.stanford.edu. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude 3.7 Sonnet (20250219) | 88.52 |
| 2 | Qwen 2 72B Instruct | 87.6 |
| 3 | Gemini 2.0 Flash Lite | 86.92 |
| 4 | GPT-4o (2024-11-20) | 86.69 |
| 5 | GPT-4o Mini (2024-07-18) | 85.6 |
| 6 | Gemini 2.0 Flash | 85.11 |
| 7 | Qwen 2.5 72B Instruct | 84.93 |
| 8 | Llama 4 Maverick Instruct FP8 | 83.98 |
| 9 | Llama 4 Scout Instruct | 83.96 |
| 10 | Llama 3.1 405B Instruct | 83.06 |
| 11 | DeepSeek V3 | 82.88 |
| 12 | Llama 3.1 70B Instruct | 82.02 |
| 13 | Llama 3.3 70B Instruct | 79.88 |
| 14 | Claude 3.5 Haiku (20241022) | 58.65 |
| 15 | Llama 3.2 90B Vision Instruct | 20.74 |
Interactive version: theaggregate.ai/benchmark?slug=helm-seahelm-nusax · How the rankings work · Data refreshed daily, snapshot 2026-07-22.