CulMind-R: leaderboard
CulMind (2026) tests multimodal models on 50 Chinese cultural heritage tasks built from more than 100 museum collections. CulMind-R is its 24-task reasoning subset; this board is the final-answer score from 0 to 100 when the model writes a structured reasoning process first; higher is better.
Metric: Final-answer score (0-100) on the 24-task CulMind-R reasoning subset, the model writing a structured reasoning process before its answer; higher is better. Source: arxiv.org. Saturation forecast: Around February 2028. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 32B Instruct | 64.9 | #585 |
| 2 | GPT-5.5 | 63.5 | #49 |
| 3 | Gemini 3 Flash (Preview) | 61.3 | #144 |
| 4 | Qwen 3.5 27B | 59.6 | #359 |
| 5 | GLM-4.5V | 58.3 | #744 |
| 6 | GPT-5.4 Mini | 54.6 | #389 |
| 7 | Qwen 3 VL 8B Instruct | 54.4 | #981 |
| 8 | Qwen 3.5 9B | 53.5 | #803 |
| 9 | InternVL3-14B | 51.3 | #1182 |
| 10 | InternVL3-8B | 50.8 | #1688 |
| 11 | Gemini 2.5 Flash | 48.2 | #453 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=culmind-r · How It Works · Data refreshed daily, snapshot 2026-11-13.