Baidu's Famou Agent 2.0 Tops MLE-Bench, Sets New SOTA for Machine Learning Engineering AI
Baidu's cloud division has announced that its enterprise-grade algorithm optimization agent, Famou Agent 2.0, has once again topped the MLE-Bench benchmark, setting a new state-of-the-art score. The formal release is planned for May 2026 at Baidu's Create AI Developer Conference.
What is MLE-Bench?
MLE-Bench is a machine learning engineering benchmark established by OpenAI, containing 75 real-world engineering problems sourced from Kaggle competitions. It tests an AI agent's ability to perform end-to-end ML engineering tasks including:
- Data preprocessing and feature engineering
- Model selection and hyperparameter tuning
- Pipeline construction
- Result optimization
Famou Agent 2.0 Details
- Developer: Baidu Smart Cloud
- Type: Enterprise-grade algorithm auto-optimization agent
- Key achievement: New SOTA on MLE-Bench (OpenAI's own benchmark)
- Formal release: May 2026 at Create 2026 Baidu AI Developer Conference
Significance
A Chinese AI company topping an OpenAI-created benchmark is significant for several reasons:
- Competitive landscape: Demonstrates that Chinese AI labs can compete with Western counterparts on rigorous engineering tasks
- Enterprise focus: The agent is designed for enterprise use, suggesting practical commercial applications
- AutoML evolution: Represents the next generation of automated machine learning, where AI agents can independently solve complex engineering problems
- Benchmark credibility: OpenAI's MLE-Bench is considered one of the most challenging tests of AI engineering capability
What's Next
Baidu will formally release Famou Agent 2.0 at Create 2026, its annual AI developer conference. The product is expected to compete with similar offerings from Western companies in the growing market for AI-powered development tools.