OMIBench - Chemistry: leaderboard

OMIBench (2026) tests vision-language models on Olympiad-level science and mathematics problems whose evidence is spread across several images. This board is the GPTScore in percent on the chemistry problems; higher is better.

Metric: GPTScore (%, binarized LLM-judge equivalence of the final answer with the gold answer, micro-averaged) on the 217 chemistry problems of OMIBench (Olympiad-level multi-image problems from biology, chemistry, mathematics and physics competitions), zero-shot; higher is better. Source: arxiv.org. Saturation forecast: Around April 2028. 37 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3 VL 235B A22B (Thinking)33.18#459
2O4 Mini32.41#368
3Qwen 3 VL 30B A3B (Thinking)32.26#719
4GPT-529.03#178
5Gemini 3 Pro (Preview)25.35#119
6InternVL3.5-241B-A28B25.35#733
7Qwen 3 VL 32B (Thinking)24.88#573
8GPT-5 Mini24.42#364
9Gemini 2.5 Pro23.96#276
10GPT-4o22.58#751
11Qwen 3 VL 235B A22B Instruct22.58#532
12InternVL3.5-38B22.12#1021
13GPT-4o Mini21.66#1667
14InternVL3.5-14B21.2#1504
15InternVL3-78B20.74#759

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=omibench-chemistry · How It Works · Data refreshed daily, snapshot 2026-11-21.