Anthropic is an AI safety and research company working to build reliable, interpretable, and steerable AI systems.

Anthropic

Claude Sonnet 4 is Anthropic's hybrid reasoning model released on May 22, 2025, representing a significant advancement in AI capabilities with enhanced coding, reasoning, and agentic performance. It serves as a balanced model offering frontier performance with practical efficiency for diverse use cases:<ul><li>Hybrid Reasoning Capabilities: Claude Sonnet 4 is a hybrid reasoning model offering two modes: near-instant responses for straightforward queries and extended thinking for deeper reasoning on complex problems, with the ability to use tools like web search during extended thinking to enhance responses.</li><li>Advanced Agentic Performance: The model demonstrates exceptional performance in agentic scenarios, it has been selected as the model powering the new coding agent in GitHub Copilot due to its superior agentic capabilities.</li><li>Enhanced Context and Safety: Claude Sonnet 4 features a 200K context window, supports up to 64K output tokens for rich code generation, and offers low rates of hallucination, making it suitable for processing large knowledge bases and complex documents. The model is 65% less likely to engage in reward hacking behavior compared to previous versions.</li><li>Constitutional AI Foundation: Claude models are built using Constitutional AI, a unique approach developed by Anthropic that trains the model to be harmless and helpful by aligning with constitutional principles including freedom, opposition to inhumane treatment, and privacy protection.</li></ul>Like all large language models of Anthropic, Claude Sonnet 4 was not specifically trained on specialized legal datasets or jurisprudential materials. While the model incorporates Constitutional AI principles that include ethical guidelines and constitutional values, this does not substitute for specialized legal training. Research demonstrates that AI systems make numerous implicit interpretive choices when analyzing legal texts, often invisibly through complex statistical computations that even experts don't fully understand. Claude Sonnet 4 may lack deep understanding of jurisdiction-specific legal frameworks, precedential reasoning, and the nuanced application of legal doctrine.

Claude Sonnet 4 is Anthropic's hybrid reasoning model released on May 22, 2025, representing a significant advancement in AI capabilities 