MM-IFEval: leaderboard
MM-IFEval (MM-IFEngine) — challenging multimodal instruction-following evaluation with compose-level and perception-level constraints, checked by rule-based verifiers and judge models; accuracy (%).
Metric: Overall Accuracy (%). Source: huggingface.co. Saturation forecast: Around January 2027. 29 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 (2025-04-14) | 72.81 | #250 |
| 2 | Gemini 2.5 Flash | 71.28 | #245 |
| 3 | Llama 4 Scout Instruct | 68.53 | #539 |
| 4 | Gemini 2.0 Flash | 67.24 | #342 |
| 5 | Claude 3.7 Sonnet | 64.38 | #249 |
| 6 | InternVL3-78B | 63.85 | #352 |
| 7 | Claude 3.5 Sonnet (20241022) | 63.7 | #291 |
| 8 | Grok 2 (1212) | 61.52 | #496 |
| 9 | InternVL2.5-78B | 58.36 | #431 |
| 10 | InternVL3-38B | 57.94 | #405 |
| 11 | Pixtral-12B | 55.25 | #808 |
| 12 | InternVL3-8B | 51.79 | #620 |
| 13 | Qwen 2.5 VL 7B | 50.99 | #643 |
| 14 | Llama 3.2 11B Instruct | 46.73 | #1126 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mm-ifeval · How It Works · Data refreshed daily, snapshot 2026-10-20.