ViMU: leaderboard
Metric: Average (%) of the four task scores (open-ended interpretation, evidence grounding, rhetoric mechanism and social value signal identification) on ViMU, short online videos whose meaning lies in subtext (irony, mockery, criticism), uniformly sampled frames, zero-shot through official implementations or APIs; questions and references written by GPT-5.4 and reviewed by human experts; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 16 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O4 Mini | 46.91 |
| 2 | Grok 4.1 Fast | 46.28 |
| 3 | Qwen 3.5 27B | 45.91 |
| 4 | GPT-5.2 | 44.67 |
| 5 | Gemini 3 Flash (Preview) | 44.31 |
| 6 | Qwen 3 VL 32B Instruct | 41.64 |
| 7 | Seed 2.0 Lite | 40.62 |
| 8 | MiMo-V2-Omni | 38.14 |
| 9 | GPT-5.4 Mini | 36.64 |
| 10 | Gemma 3 27B (IT) | 36.43 |
| 11 | Ministral 3 14B | 35.45 |
| 12 | Ministral 8B | 34.79 |
| 13 | GLM-4.5V | 25.94 |
| 14 | Gemma 3 4B (IT) | 23.28 |
| 15 | Claude 3 Haiku | 22.9 |
Interactive version: theaggregate.ai/benchmark?slug=vimu · How It Works · Data refreshed daily, snapshot 2026-10-07.