FrontierScience

A benchmark for research-level scientific reasoning, designed to separate frontier models on difficult science tasks that mix domain knowledge with deep reasoning.

1 models

Higher is better.

1GPT-5.4 ProOpenAI36.7

Higher is better. Bar colour: blue proprietary, orange open weights, green other.

About this benchmark

Metric: Scientific reasoning benchmark
Tasks: Research-level science tasks
Authors: OpenAI (2026)
Source: FrontierScience ↗

Scores are as reported by model developers or evaluators and may use different settings (for example reasoning effort or tools). Compare with care.