#320 · Primary category: AI Agents & Automation

MiroThinker

agent agent-framework browsecomp deep-research futurex gaia hle research-agent search-agent xbench

MiroThinker is a deep research agent optimized for complex research and prediction tasks. Our latest models, MiroThinker-1.7, achieves 74.0 and 75.3 on the BrowseComp and BrowseComp Zh, respectively.

Project last updated:07/06/26

GitHub Stars

8.4K

Forks

644

Contributors

14

License

Apache-2.0

Why we included this project

Instead of answering straight from memory, MiroThinker searches the web and reads documents in several formats, then iterates through many tool calls before assembling a cited, structured report. That makes it a good fit for market scans or technical investigations where an open model you can run yourself is the point. The repo ships trained weights at a few sizes, deployment workflows, and a demo, so you can stand up your own research pipeline without building the agent harness from scratch. The step-by-step trace collection is also handy if you want to verify how a conclusion was reached rather than trusting the output blindly.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category