On January 24, Baichuan Intelligence officially launched the first all-scenario deep thinking model in China — Baichuan-M1-preview. This model demonstrates outstanding performance in three dimensions: language, vision, and search reasoning, marking another significant breakthrough in the field of AI.
Baichuan-M1-preview has excelled in multiple authoritative evaluation sets, particularly in mathematical reasoning and coding tasks, surpassing many well-known models, including o1-preview. Notably, its visual reasoning capabilities have outperformed top models such as GPT-4o and Claude3.5Sonnet in assessments like MMMUVa and MathVista. This model has now officially launched in the Baixiao Ying app, endowing the application with deep thinking capabilities.
In the medical field, Baichuan-M1-preview showcases its unique advantages. By establishing a rigorous medical reasoning process, it can conduct in-depth analyses of complex medical issues and provide decision support for doctors. Baichuan Intelligence's self-built medical knowledge base covers hundreds of millions of entries, dynamically updating the latest medical research and authoritative guidelines to ensure the accuracy and reliability of information.
It is worth mentioning that the model's "evidence-based medical mode" enhances its authority in addressing medical issues, providing well-founded answers through evidence-based medicine methods. For professionals, Baichuan-M1-preview can quickly access the latest research findings, improving clinical decision-making efficiency. For patients, this model allows them to obtain scientifically validated medical information and advice, enhancing the quality of communication with their doctors.
Baichuan Intelligence has also open-sourced a smaller version of Baichuan-M1-14B, making it the industry's first medical-enhanced open-source model. Its performance is excellent, with results comparable to larger parameter models. Through innovative data collection and training methods, the Baichuan-M1 series stands out in reasoning complex medical problems, paving new directions for AI applications in the medical field.
