OpenVLM AI2D - Rock Strata: leaderboard

OpenCompass OpenVLM AI2D subcategory score for diagram questions about rock strata.

Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 284 models tracked.

Top models

#ModelScoreOverall rank
1InternVL3-78B95.1#352
2Qwen 2 VL 72B92.7#460
3GPT-5 Mini (2025-08-07)92.7#172
4GPT-4.592.7#274
5Gemini 2.5 Pro90.2#152
6GPT-590.2#98
7InternVL2.5-78B90.2#431
8InternVL3-38B87.8#405
9GPT-4.1 (2025-04-14)87.8#250
10GPT-4o ChatGPT87.8
11GPT-5 Nano85.4#425
12InternVL3-14B85.4#505
13Qwen 2.5 VL 7B85.4#643
14Claude 3.7 Sonnet82.9#249
15Gemini 1.0 Pro82.9#933

No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.

Interactive version: theaggregate.ai/benchmark?slug=openvlm-ai2d-rock-strata · How It Works · Data refreshed daily, snapshot 2026-10-20.