MIBench-mini (Multimodal Interaction): leaderboard
Metric: Overall accuracy (%) on MIBench-mini, the paper's subset on which the closed-source models were also evaluated (multiple-choice image-text tasks; vision- and text-centric samples average several context variations); higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 29 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 79.97 | #64 |
| 2 | Gemini 2.5 Pro | 76.6 | #145 |
| 3 | O3 (2025-04-16) | 73.53 | #117 |
| 4 | GPT-5.1 (2025-11-13) | 72.73 | #97 |
| 5 | GPT-5.2 | 71.6 | #105 |
| 6 | Claude Sonnet 4.5 | 70.8 | #138 |
| 7 | Qwen 3 VL 235B A22B Instruct | 66.67 | #264 |
| 8 | Qwen 2.5 VL 72B Instruct | 63.87 | #364 |
| 9 | GPT-4o (2024-11-20) | 62.2 | #369 |
| 10 | Qwen 2.5 VL 32B Instruct | 60.67 | #443 |
| 11 | Qwen 3 VL 8B Instruct | 58.53 | #401 |
| 12 | InternVL3-8B | 56.47 | #606 |
| 13 | Qwen 2 VL 7B Instruct | 55.4 | #816 |
| 14 | Qwen 2.5 VL 7B Instruct | 54.67 | #643 |
| 15 | InternVL2-8B | 51.2 | #826 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=mibench-mini-multimodal-interaction · How It Works · Data refreshed daily, snapshot 2026-10-11.