You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
-**[2025-09-15]**: 🎉🎉 **MiroFlow v0.3** - Enhanced codebase architecture and significantly improved benchmark performance. MiroFlow now ranks #1 in the future prediction benchmark.
59
-
-**[2025-08-27]**: **MiroFlow v0.2** - Achieves state-of-the-art performance across [multiple agentic benchmarks](https://miromind.ai/blog/miroflow), including HLE (27.2%), HLE-Text-Only (29.5%), BrowserComp-EN (33.2%), BrowserComp-ZH (47.1%), and xBench-DeepSearch (72.0%)
60
-
-**[2025-08-26]**: Released [GAIA Validation Trace](docs/public_trace.md) (73.94% pass@1) and [Gradio Demo](https://github.com/MiroMindAI/MiroThinker/tree/main/apps/gradio-demo) for local deployment
61
-
-**[2025-08-08]**: 🎉 **MiroFlow v0.1** - Complete open-source release of framework, models, and training data
**MiroFlow** is a comprehensive framework for building intelligent AI agents that achieve state-of-the-art performance on complex reasoning tasks. It provides enhanced conversation management, flexible tool integration, and extensive benchmark evaluations across multiple datasets.
68
-
69
-
**MiroThinker** is the open-source agentic model series built on this framework.
35
+
This repo is the official implementation of the MiroMind Research Agent Project. It is a leading-performance, fully open-source system designed to perform multi-step internet research for addressing complex challenges such as future event prediction. The project currently comprises four key components:
70
36
71
-
### 🌟 Key Highlights
37
+
- 🤖 **MiroFlow**: an open-source research agent framework that offers reproducible state-of-the-art performance on representative benchmarks (e.g., FutureX, GAIA, HLE, xBench-DeepSearch, and BrowserComp benchmarks), included in this repo. See [[Get Started in Under 5 Minutes]](#-get-started-in-under-5-minutes) for a quick start.
38
+
- 🤔 **MiroThinker**: an open-source agent foundation model that natively supports tool-assisted reasoning. See [MiroThinker](https://github.com/MiroMindAI/mirothinker).
39
+
- 📊 **MiroVerse**: 147k premium open-source training data supporting research agent training. See [MiroThinker](https://huggingface.co/datasets/miromind-ai/MiroVerse-v0.1).
40
+
- 🚧 **MiroTrain / MiroRL**: The training infra that supports stable and efficient training for the research agent models. See [MiroTrain](https://github.com/MiroMindAI/MiroTrain) / [MiroRL](https://github.com/MiroMindAI/MiroRL)
72
41
73
-
- 🏆 **State-of-the-Art Performance**: #1 ranking across [multiple agentic benchmarks](https://miromindai.github.io/MiroFlow/evaluation_overview/)
74
-
- 📊 **Premium Training Data**: Curated datasets via [MiroVerse](https://huggingface.co/datasets/miromind-ai/MiroVerse-v0.1)
75
-
- 🤖 **Open Models**: Complete collection at [MiroThinker](https://huggingface.co/collections/miromind-ai/mirothinker-v01-689301b6d0563321862d44a1)
76
-
- 🔧 **Full Training Stack**: SFT/DPO recipes at [MiroTrain](https://github.com/MiroMindAI/MiroTrain)
77
-
- 🎯 **Advanced RL**: Reinforcement learning via [MiroRL](https://github.com/MiroMindAI/MiroRL)
-**[2025-09-15]**: 🎉🎉 **MiroFlow v0.3**: Enhanced codebase architecture and significantly improved benchmark performance, boosting GPT-5's prediction accuracy for future events by 11%.
62
+
MiroFlow now ranks #1 in the future prediction benchmark. See [FutureX](https://futurex-ai.github.io/).
63
+
-**[2025-08-27]**: **MiroFlow v0.2**: Achieves state-of-the-art performance across [multiple agentic benchmarks](https://miromind.ai/blog/miroflow), including HLE (27.2%), HLE-Text-Only (29.5%), BrowserComp-EN (33.2%), BrowserComp-ZH (47.1%), and xBench-DeepSearch (72.0%).
64
+
-**[2025-08-26]**: Released [GAIA Validation Trace](docs/public_trace.md) (73.94% pass@1) and [Gradio Demo](https://github.com/MiroMindAI/MiroThinker/tree/main/apps/gradio-demo) for local deployment.
65
+
-**[2025-08-08]**: **MiroFlow v0.1**: Complete open-source release of framework.
102
66
103
-
# 🚀 Get Started in Under 5 Minutes
67
+
---
104
68
105
-
Clone the repository, configure your API key, and run your first intelligent agent. You'll just need one `OPENROUTER_API_KEY`.
**Example**: Intelligent document analysis with file processing capabilities.
116
80
@@ -131,30 +95,86 @@ uv run main.py trace --config_file_name=agent_quickstart_1 --task="What is the f
131
95
132
96
> **💡 Tip:** If you encounter issues, check that your API key is correctly set in the `.env` file and that all dependencies are installed.
133
97
134
-
**🎯 Comprehensive Benchmark Suite**:
135
-
-**GAIA Validation**: A benchmark for General AI Assistants. ([paper](https://arxiv.org/abs/2311.12983))
136
-
-**GAIA-Text-103**: A subset of GAIA Validation for text-only tasks. ([paper](https://arxiv.org/abs/2505.22648))
137
-
-**HLE**: Humanity's Last Exam. ([paper](https://arxiv.org/abs/2501.14249))
138
-
-**HLE-Text-500**: A subset of HLE for text-only tasks. ([paper](https://arxiv.org/pdf/2504.21776))
139
-
140
-
Follow our detailed guides to reproduce benchmark results in our [Benchmarks Documentation](https://miromindai.github.io/MiroFlow/evaluation_overview/)
98
+
---
141
99
142
-
# 🤖 MiroFlow: AI Agentic Foundation Framework
100
+
##🤖 What is MiroFlow?
143
101
144
-
MiroFlow is a high-performance, modular framework for building intelligent AI agents that deliver state-of-the-art results on complex reasoning tasks. The framework features advanced multi-turn conversation capabilities, extensive tool ecosystem integration, and hierarchical sub-agent orchestration for optimal task completion. Learn more about our agent [workflow architecture](https://miromindai.github.io/MiroFlow/core_concepts/).
102
+
MiroFlow is a high-performance, modular framework for building intelligent AI agents that deliver state-of-the-art results on complex reasoning tasks like future event prediction. The framework features advanced multi-turn conversation capabilities, extensive tool ecosystem integration, and hierarchical sub-agent orchestration for optimal task completion. Learn more about our [agent framework](https://miromindai.github.io/MiroFlow/core_concepts/).
-**Reproducible State-of-the-Art Performance**: #1 ranking across [multiple agentic benchmarks](https://miromindai.github.io/MiroFlow/evaluation_overview/), including FutureX, GAIA, HLE, xBench-DeepSearch, and BrowserComp benchmarks)
127
+
-**High Concurrency & Reliability**: Built with robust concurrency management and fault-tolerant design, MiroFlow efficiently handles rate-limited APIs and unstable networks, ensuring seamless trajectory collection and reliable execution of complex tasks.
128
+
-**Cost-Effective Deployment**: Powered by the open-source MiroThinker model, MiroFlow can run a research agent service on a single RTX 4090. The entire stack relies on free, open-source tools, making it simple to deploy, scale, and reproduce.
We achieved the #1 ranking on the FutureX Benchmark Leaderboard as of September 10, 2025, boosting GPT-5's prediction accuracy for future events by 11%.
0 commit comments