M3 Hate Memes (Meme Only) - Category Macro-F1: leaderboard

Metric: Macro-F1 (%) over the eight hate categories (religion, politics, race, gender, health status, violence, public health, international relations) of multi-label category classification on M3's 1,318 hate memes; the model sees only the meme image (with its embedded text); zero-shot; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 8 models tracked.

Top models

#ModelScoreOverall rank
1GLM-4.1V-9B (Thinking)69.02#457 (GLM-4.1V-9B)
2Qwen 3 VL 8B Instruct68.17#401
3Qwen 2.5 VL 7B Instruct65.54#643
4Gemini 3 Pro64.6#77
5GPT-4o60.93#333

No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.

Interactive version: theaggregate.ai/benchmark?slug=m3-hate-memes-meme-only-category-macro-f1 · How It Works · Data refreshed daily, snapshot 2026-10-11.