HELM SeaHELM - MLHSD: leaderboard
HELM SeaHELM Southeast-Asian-languages evaluation: Indonesian multi-label hate-speech detection.
Metric: Macro F1 score. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.7 Sonnet (20250219) | 73.81 | #201 |
| 2 | GPT-4o (2024-11-20) | 72.31 | #381 |
| 3 | Llama 4 Maverick Instruct FP8 | 68.92 | #414 |
| 4 | Gemini 2.0 Flash | 68.55 | #342 |
| 5 | Llama 3.1 405B Instruct | 67.96 | #457 |
| 6 | Llama 3.3 70B Instruct | 66.39 | #534 |
| 7 | Qwen 2 72B Instruct | 66.01 | #576 |
| 8 | Llama 3.1 70B Instruct | 65.32 | #561 |
| 9 | Llama 4 Scout Instruct | 64.59 | #539 |
| 10 | GPT-4o Mini (2024-07-18) | 64.44 | #558 |
| 11 | Qwen 2.5 72B Instruct | 63.02 | #454 |
| 12 | Gemini 2.0 Flash Lite | 61.75 | #450 |
| 13 | Llama 3.1 8B CPT SEA LIONv3 Instruct | 53.92 | #810 |
| 14 | Gemma 2 27B (IT) | 43.56 | #645 |
| 15 | Gemma 2 9B (IT) | 27.54 | #788 |
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-mlhsd · How It Works · Data refreshed daily, snapshot 2026-10-20.