AIMultiple - TweetEval Hate: leaderboard
AIMultiple zero-shot TweetEval Hate: flag hate speech in the first 200 test tweets of the TweetEval hate task with a single-label reply at temperature 0. Scored with macro-F1 as defined by TweetEval and published as whole numbers; %, higher is better.
Metric: Macro-F1 (%). Source: aimultiple.com. Saturation forecast: Not forecast. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 82 | #94 |
| 2 | GPT-5.5 | 80 | #33 |
| 3 | Qwen 3.6 Plus | 77 | #124 |
| 4 | Kimi K2.6 | 76 | #113 |
| 5 | GLM-5.1 | 76 | #109 |
| 6 | MiniMax-M2.7 | 75 | #270 |
| 7 | Gemini 3.1 Pro (Preview) | 71 | #64 |
| 8 | Claude Opus 4.8 | 71 | #39 |
| 9 | Gemini 3.5 Flash | 65 | #66 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=aimultiple-tweeteval-hate · How It Works · Data refreshed daily, snapshot 2026-11-05.