ChatLaw

ChatLaw is developed by Peking University Shenzhen Graduate School and Peng Cheng Laboratory. It employs a multi-model collaborative design: Keyword LLM extracts legal keywords, Law LLM retrieves relevant statutes and precedents, and ChatLaw LLM generates natural language answers. Users can chat for legal guidance, upload files or audio to auto-extract key facts, and receive mind maps and charts, making legal information more accessible.

Its strengths lie in being open-source, fine-tuned for Chinese legal scenarios (e.g., using judgment datasets), and supporting multimodal input. Two model sizes are available—ChatLaw-13B (based on Ziya-LLaMA, solid Chinese performance) and ChatLaw-33B (based on Anima, stronger reasoning but sometimes responds in English due to insufficient Chinese pre-training).

Key drawbacks include incomplete open-source code (community criticism), a demo site (chatlaw.cloud) that has suffered SSL issues and login failures, signaling maintenance and reliability risks. The AGPL-3.0 license requires that any network service using the model must release the full modified source code, which limits commercial deployment.