OmniMatBench - Calculation: leaderboard

OmniMatBench (2026) poses 3,171 expert-curated multimodal materials-science problems, open-ended questions graded against expert key points and calculations graded slot by slot, across 19 subfields. This board is the strict exact-slot accuracy of the calculation answers, macro-averaged over the subfields; percent, higher is better.

Metric: Calculation strict exact-slot accuracy (%): share of structured answer slots (values, units, formulas) exactly matching the reference after normalization, macro-averaged over the 19 subfields, on the 1,910 calculation questions on OmniMatBench (3,171 expert-curated multimodal materials-science problems in 19 subfields: 1,261 open-ended QA and 1,910 calculation questions); higher is better. Source: arxiv.org. Saturation forecast: Around October 2028. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Claude Opus 4.729.1#85
2Gemini 3.1 Pro (Preview)28.7#99
3GPT-5.525.5#54
4Qwen 3.6 Plus24.9#238
5Intern-S1-Pro23.2#585
6Grok 4.322.9#342
7Qwen 3 VL 235B A22B20.7#485
8O4 Mini20.1#368
9O319.9#237
10Llama 4 Maverick19.3#1110
11Gemma 4 31B19.2#382
12GLM-4.6V17.4#710
13GPT-4o7.2#751

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

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