Creation-MMBench: leaderboard
Creation-MMBench (OpenCompass) — context-aware creative writing and content generation for multimodal models across image-grounded real-world scenarios; judged against a GPT-4o reference, reward score (-100 to 100).
Metric: Overall Reward. Source: huggingface.co. Saturation forecast: Estimated already saturated. 29 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Flash | 6.61 | #245 |
| 2 | GPT-4.1 (2025-04-14) | 5.61 | #250 |
| 3 | InternVL3-78B | -6.87 | #352 |
| 4 | Grok 2 (1212) | -8.86 | #496 |
| 5 | Gemini 2.0 Flash | -8.9 | #342 |
| 6 | InternVL3-38B | -8.99 | #405 |
| 7 | Claude 3.7 Sonnet | -11.28 | #249 |
| 8 | InternVL3-8B | -14 | #620 |
| 9 | Claude 3.5 Sonnet (20241022) | -15.46 | #291 |
| 10 | Llama 4 Scout Instruct | -16.44 | #539 |
| 11 | InternVL2.5-78B | -16.44 | #431 |
| 12 | Pixtral-12B | -21.45 | #808 |
| 13 | Qwen 2.5 VL 7B | -29.97 | #643 |
| 14 | Llama 3.2 11B Instruct | -48.23 | #1126 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=creation-mmbench · How It Works · Data refreshed daily, snapshot 2026-10-20.