Open Portuguese LLM - Hate Speech: leaderboard
Portuguese hate-speech detection task from the Open Portuguese LLM Leaderboard, reporting macro-F1 on Brazilian Portuguese tweets annotated for hate speech.
Metric: F1 (%). Source: huggingface.co. Saturation forecast: Around May 2028. 42 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.7 Sonnet (20250219) | 76.99 | #201 |
| 2 | Qwen 2.5 VL 72B Instruct | 76.96 | #377 |
| 3 | Qwen 2.5 VL 32B Instruct | 76.96 | #453 |
| 4 | Gemini 2.0 Flash (001) | 76.55 | #360 |
| 5 | Qwen 2.5 72B Instruct | 76.02 | #454 |
| 6 | Kimi K2 | 75.36 | #243 |
| 7 | GPT-4o (2024-08-06) | 75.13 | #336 |
| 8 | Gemini 2.5 Flash Lite | 75.12 | #422 |
| 9 | Gemini 2.5 Flash | 75.03 | #245 |
| 10 | GPT-5 (Minimal) | 75.02 | #98 (GPT-5) |
| 11 | GPT-4o Mini (2024-07-18) | 75.01 | #558 |
| 12 | Gemini 2.0 Flash Lite (001) | 75.01 | #544 |
| 13 | GPT-5 Mini (2025-08-07) | 74.77 | #172 |
| 14 | DeepSeek V3.1 | 74.23 | #264 |
| 15 | Gemini 1.5 Flash (002) | 74.06 | #583 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=open-portuguese-llm-hate-speech · How It Works · Data refreshed daily, snapshot 2026-10-20.