Baichuan's Medical AI Breakthrough Surpasses Human Doctors

In a development that could reshape healthcare diagnostics, Baichuan Intelligence today launched its Baichuan-M3 model - an open-source medical AI that outperformed both human physicians and OpenAI's latest GPT-5.2 across multiple benchmarks.

Benchmark Dominance

The new model topped four critical evaluations:

• Diagnostic accuracy
• Medical hallucination control
• Healthbench standard tests
• Healthbench Hard challenges

What makes these results remarkable? The system didn't just edge past competitors - it established clear superiority across all testing scenarios.

"When we started in 2023, skepticism was high," recalled founder Wang Xiaochuan during the Beijing launch event. "But after DeepSeek's 2025 release, the medical community began seeing AI differently. Today's results confirm that shift."

The comparison to Baidu drew chuckles from attendees when Wang noted: "People find this significantly more reliable than existing alternatives."

Healthcare's AI Tipping Point

With tech giants like OpenAI and Anthropic doubling down on medical applications, Wang positioned Baichuan-M3 as arriving at healthcare's digital inflection point.

"We're past the theoretical phase," he emphasized. "Medical AI isn't coming - it's here, delivering real clinical value today."

The open-source approach could accelerate adoption, allowing hospitals and researchers worldwide to implement and build upon Baichuan-M3 without licensing barriers.

Key Points:

• Performance leap: Baichuan-M3 exceeds GPT-5.2 and human benchmarks
• Open access: Model released as open-source to spur innovation
• Industry shift: Growing acceptance of diagnostic AI among physicians
• Practical focus: Designed specifically for real-world clinical applications