What is Baichuan-M1-preview
Baichuan-M1-preview is the first domestic full-scenario deep thinking model launched by Baichuan Intelligence. The model has reasoning capabilities in the three major domains of language, vision, and search, and performs excellently in multiple authoritative evaluations such as mathematics and code, surpassing models such as o1-preview. Its core highlight is that it unlocks the "medical evidence-based mode," which, through a self-built evidence-based medicine knowledge base with hundreds of millions of entries, can quickly and accurately answer medical clinical and scientific research questions.
Main features of Baichuan-M1-preview
- Multi-domain reasoning capability: Baichuan-M1-preview simultaneously has reasoning capabilities in the three major domains of language, vision, and search.
- Language reasoning: Its results on mathematical benchmarks such as AIME and Math as well as on the LiveCodeBench code task surpass o1-preview.
- Visual reasoning: In authoritative evaluations such as MMMU-val and MathVista, its performance is better than leading domestic and international models such as GPT-4o, Claude3.5 Sonnet, and QVQ-72B-Preview.
- Search reasoning: It has strong information retrieval and integration capabilities.
- Medical evidence-based mode: The model unlocks the "medical evidence-based mode." Through a self-built evidence-based medicine knowledge base with hundreds of millions of entries, it can quickly and accurately answer medical clinical and scientific research questions. It realizes a complete end-to-end service from evidence retrieval to deep reasoning, helping users make the best decisions when facing complex medical problems.
- Deep thinking mode: In deep thinking mode, Baichuan-M1-preview can accurately answer questions such as mathematics, code, and logical reasoning. Like a senior medical expert, through deep thinking it constructs a rigorous medical reasoning process, providing users with comprehensive disease analysis and personalized health management suggestions.
Technical principles of Baichuan-M1-preview
- Data and training methods: In order to improve the model's medical reasoning capability, Baichuan Intelligence collected trillions of tokens of serious medical data and generated over one hundred billion tokens of diversified data, covering medical complex decision reasoning chains, decision bases, and more. During the training stage, the model adopts a multi-stage domain improvement plan and introduces the ELO reinforcement learning method to optimize the chain-of-thought path, improving generation quality and logical reasoning capability.
- Open-source version support: The small-size version of Baichuan-M1-preview, Baichuan-M1-14B, as the industry's first open-source medical-enhanced large model, has medical capabilities surpassing the larger-parameter Qwen2.5-72B and comparable to o1-mini.
How to use Baichuan-M1-preview
- Visit the platform: Baixiaoying has officially launched, and you can log in to visit Baixiaoying'sofficial websiteto experience it.
- Select deep thinking: Turn on deep thinking mode to answer questions such as mathematics, code, and logical reasoning.
Application scenarios of Baichuan-M1-preview
- Medical field: Baichuan-M1-preview unlocks the "medical evidence-based mode." Through a self-built evidence-based medicine knowledge base with hundreds of millions of entries, it can quickly and accurately answer medical clinical and scientific research questions. Like a senior medical expert, through deep thinking it constructs a rigorous medical reasoning process, providing users with comprehensive disease analysis and personalized health management suggestions.
- Education field: The model performs excellently in mathematics and code evaluations, can provide learning assistance to students, and help them better understand and master complex concepts.
- Scientific Research and Data Analysis: The multi-domain reasoning capability of Baichuan-M1-preview enables it to handle complex scientific research problems, and in medical research, it can greatly shorten the time for research exploration.
- Content Creation and Customer Service: The model can generate high-quality text content and support work such as content creation and copywriting. Its language reasoning capability gives it application potential in the customer service field, enabling it to quickly and accurately answer user questions.
- Creative Industry: The multi-domain reasoning capability of Baichuan-M1-preview enables it to process text and visual information simultaneously, providing new ideas and tools for creative work, such as generating creative copy and parsing image content.