Welcome to MiroFlow
🚀 What is MiroFlow?
MiroFlow is an agentic AI platform for building intelligent agents with flexible tool integration and comprehensive benchmark evaluations.
📝 Recent Updates
Latest Changes & Improvements
Oct 2025 -
Week 2
- 🌐 Added support for BrowseComp-ZH (Chinese) evaluation benchmark #88
- 🤖 Added support for DeepSeek V3.1 with OpenRouter and tool call format #86
- 🚶 Added support for WebWalkerQA dataset evaluation #84
- 📖 Updated documentation for reading and audio tools #83
- 📝 Added support for HLE-Text evaluation benchmark #81
- 🧠 Added support for HLE (Humanity's Last Exam) benchmark #79
- 🌐 Added support for BrowseComp-EN evaluation benchmark #78
- 🔌 Added support for MiroAPI integration #76
Week 1
- 📊 Added support for FinSearchComp evaluation benchmark #51
- 🔍 Added support for XBench-DS (Deep Search) evaluation #47
- 🧠 Updated o3 hints and summary to more models #58
- ✨ Added support for GPT-5 integration #52
- 🔧 Improved tool logs and per-task log storage #69
- 🤖 Added support for single agent mode #67
- 📚 Added comprehensive collection of agentic AI research papers #65
🎯 Quick Start
Ready to get started? Choose your path:
Get Started in Minutes
Jump right in with our quickstart guide:
Evaluate existing models on benchmarks:
Build your own AI agents:
🔗 Ecosystem
Explore the complete MiroMind AI ecosystem:
MiroMind AI Products
| Name | Description | Link |
|---|---|---|
| MiroFlow | Core framework for AI agent platform | Documentation |
| MiroThinker | State-of-the-art agent foundation models | Hugging Face |
| MiroVerse | Curated datasets for model training | Dataset |
| MiroTrain | Complete training recipes and tools | GitHub |
Documentation Info
Last Updated: September 2025 · Doc Contributor: Team @ MiroMind AI · Version: v0.3