ChatLaw
Product information, use cases, and access for ChatLaw.
Pricing information
No verified public pricing is available yet.
DevPrice organizes public information and does not sell ChatLaw subscriptions. Prices and availability are determined by ChatLaw.
What is ChatLaw?
ChatLaw is an open-source Chinese legal large language model and multi-agent legal assistant developed by a Peking University team. It combines large language models with knowledge graphs and a mixture-of-experts architecture for work that depends on understanding Chinese legal language, including legal questions and analysis.
Its workflow uses multiple agents to gather relevant information, conduct legal research, and organize the resulting consultation into a report. Legal professionals, students, and researchers can use it to explore legal knowledge and analyze cases, while developers can study its open-source models and build related applications.
Key features of ChatLaw
Multi-Agent Consultation
ChatLaw divides information gathering, legal research, synthesis, and report preparation into collaborating stages that reflect a structured legal-workflow approach. This is suited to complex legal questions that benefit from being organized into a coherent consultation result.
Mixture-of-Experts Models
ChatLaw2-MoE is based on the InternLM architecture and a 4x7B mixture-of-experts design, adapted for Chinese legal language processing. The project also provides ChatLaw-13B and ChatLaw-33B demo models, giving users different model forms for general tasks and more demanding legal question answering.
Legal Text Similarity Search
ChatLaw-Text2Vec is a text-similarity model trained on 93,000 court decisions. It can match a user query with relevant legal provisions or case materials, helping legal research narrow its candidates and identify contextually related references for further reading.
Legal Tasks and Knowledge Data
The project covers tasks such as legal question answering, case classification, keyword extraction, and logical reasoning, supported by legal-domain datasets and external knowledge resources. This combination is useful for studying legal cognition tasks, analyzing legal text, and exploring model applications for Chinese legal settings.