Baichuan Unveils Baichuan-M1-Preview: A Full-Scenario Deep Reasoning Model 130

Baichuan Intelligence released China's first open-source large language model specifically optimized for medical scenarios, built from scratch, on January 24: Baichuan-M1-Preview. This model is the only one in China that simultaneously possesses reasoning capabilities across three major domains: language, vision, and search, marking a significant breakthrough in artificial intelligence technology. Baichuan-M1-Preview not only performs excellently in general capabilities but also delivers a complete service from evidence retrieval to in-depth reasoning in the medical field.

Features of the Baichuan-M1-Preview Model

• Multimodal Reasoning Capability: The model can process linguistic, visual, and search information, providing comprehensive reasoning support.

• Medical Evidence-Based Mode: The model has specially unlocked a medical evidence-based mode, enabling rapid retrieval and in-depth reasoning from massive amounts of medical information. This feature makes the model's application in the medical field particularly outstanding, helping doctors provide a scientific basis for clinical decision-making.

• Efficient Reasoning Capability: In multiple authoritative benchmarks, Baichuan-M1-Preview outperforms many existing models, especially in mathematical and code reasoning, demonstrating powerful capabilities.

Application Scenarios of Baichuan-M1-Preview

• Medical Field: Through the medical evidence-based mode, the model can help doctors quickly access the latest research findings and authoritative information, improving the accuracy and efficiency of clinical diagnosis.

• Education and Training: In the education sector, the model can be used to assist teaching, provide personalized learning suggestions and answers, and help students better understand complex concepts.

• Research Support: Researchers can leverage this model for literature retrieval and data analysis, quickly obtaining relevant research results and improving research efficiency.

• Enterprise Decision-Making: In business environments, the model can help enterprises analyze market trends, customer needs, and other information to support decision-making.

Baichuan-M1-Preview also possesses powerful visual reasoning capabilities, with results in authoritative benchmarks such as MMMU-val and MathVista surpassing top domestic and international models like GPT-4o and Claude 3.5 Sonnet. This capability enables the model to perform excellently when handling images and visual information, allowing for accurate recognition and judgment in complex scenarios.

In addition, the medical evidence-based mode of Baichuan-M1-Preview is supported by a self-built evidence-based medical knowledge base containing hundreds of millions of entries, covering a large amount of medical papers, authoritative guidelines, and expert consensus, ensuring that the model can provide reasoning and recommendations based on the latest medical evidence. This feature not only enhances the scientific rigor of medical decision-making but also provides more reliable support for doctors and patients.

In terms of enterprise decision-making, Baichuan-M1-Preview can help enterprises make more precise strategic decisions in market competition through deep thinking and data analysis, improving operational efficiency and market responsiveness. Its multimodal nature allows the model to integrate information from different fields, providing users with comprehensive solutions.

Github:https://github.com/baichuan-inc/Baichuan-M1-14B

Huggingface(base):https://huggingface.co/baichuan-inc/Baichuan-M1-14B-Base

Huggingface(Instruct):https://huggingface.co/baichuan-inc/Baichuan-M1-14B-Instruct