MatOS has been officially open-sourced.
The story so far
Beijing Zhongguancun College, Zhongguancun Artificial Intelligence Research Institute have open-sourced the materials science intelligent agent MatOS, which integrates 3,135 datasets, 386 skills, 104 workflows, and 51 expert roles, and integrates 48 material calculation software. MatOS has completed a variety of tasks from end-to-end autonomous research, tool calls to large-scale screening, recipe design, and new method development in 3 intelligent agent benchmarks and 8 research cases. Compared to Codex CLI using the same base model on ResearchClawBench, it achieved an average score 33.6% higher and an 80.8% higher score on material tasks; the success rate of complete problems on MatTools was 97.96%. The technical report has been published on arXiv, and the code has been open-sourced on GitHub.
AI-written from the reports · Updated 56 minutes ago
Coverage timeline
Follow the coverage to see every side of the story.
- QbitAIBeijing Zhongguancun College and Zhongguancun AI Research Institute have released the material science agent MatOS
MatOS completes a variety of tasks including end-to-end autonomous research, tool invocation, large-scale screening, formulation design, and new method development. Compared to Codex CLI using the same base model, it achieves an average score 33.6% higher and 80.8% higher on material tasks on ResearchClawBench; on MatTools, it achieves a complete problem success rate of 97.96%, integrating 3,135 datasets, 386 skills, 104 workflows, and 51 expert roles, with the code open-sourced on GitHub.
Heat over time
Not enough data yet to draw a trend.