PolyReal - Performance and Application Exploration: leaderboard

PolyReal (2026) evaluates multimodal models on 545 expert-written polymer science tasks covering five stages of real experimental work. Answers are scored against expert key points by an LLM judge and summarized as F1 from 0 to 1. This board is F1 on the module about judging structure-property relations and suitability for an application; higher is better.

Metric: F1 score on the Performance and Application Exploration module of PolyReal (structure-property reasoning and material suitability); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-50.623#178
2O30.609#237
3Gemini 2.5 Pro0.555#276
4Grok 40.54#319
5Claude Sonnet 4.5Thinking0.539#261
6Qwen 3 VL 32B (Thinking)0.538#573
7Qwen 3 VL 235B A22B (Thinking)0.518#459
8Gemini 2.5 Flash0.495#463
9Intern-S10.476#624
10Gemini 2.0 Flash (Thinking)0.468
11DeepSeek R10.427#482
12Qwen 2.5 VL 72B Instruct0.402#841
13GPT-4o0.37#751
14GPT-4o Mini0.351#1667
15DeepSeek-VL20.122#2927

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

Interactive version: theaggregate.ai/benchmark?slug=polyreal-performance-and-application-exploration · How It Works · Data refreshed daily, snapshot 2026-11-21.