MemeBench - Visual Clues: leaderboard
Metric: Visual-clue coverage (%) on 1,253 Chinese and English memes centred on anime, comics, games and online subcultures, explained closed-book from the image; scored against VIKR checklists under the intersection of Gemini-3.1-Pro and GPT-5.1 judges, unscored responses counted as failures. Source: arxiv.org. Saturation forecast: Around December 2026. 26 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 92.3 |
| 2 | Gemini 3 Flash | 86 |
| 3 | Claude Opus 5 | 74.7 |
| 4 | GPT-5.6 Sol | 72.5 |
| 5 | Kimi K2.5 | 71.1 |
| 6 | Qwen 3.6 Plus | 65.1 |
| 7 | GPT-5.6 Terra | 61.9 |
| 8 | Claude Sonnet 5 | 59.2 |
| 9 | Qwen 3 VL 235B A22B Instruct | 53.2 |
| 10 | Qwen 3.6 Flash | 53.2 |
| 11 | GPT-5.6 Luna | 52.1 |
| 12 | Qwen 3 VL 235B A22B (Thinking) | 50.2 |
| 13 | MiMo-V2-Omni | 48.1 |
| 14 | Qwen 3 VL 32B Instruct | 47.6 |
| 15 | MiMo-V2.5 | 47.2 |
Interactive version: theaggregate.ai/benchmark?slug=memebench-visual-clues · How It Works · Data refreshed daily, snapshot 2026-09-29.