anthropic released clause 3.7 Sonnet, a new version of its language model, which unites quick answers with in -depth thought. Users can manage the time that the model spends on analysis, and the API allows you to set limits on calculations.
One of the key areas of improvement was programming. Together with the model, they introduced Claude Code – a tool for automation of development. He can analyze the code, edit files, launch tests, work with GitHub repositories and execute commands in the terminal.
Claude 3.7 Sonnet is available in all anthropic tariffs and API platforms, including Amazon Bedrock and Google Cloud Vertex ai.
Model capabilities:
– in standard mode works as an improved version of Claude 3.5 Sonnet.
– in in-depth thought mode, increases the accuracy of answers in complex areas: programming, mathematics, physics.
– API allows you to limit the number of tokens spent on analysis, balancing the speed and quality of the answer.
– knows how to work with GitHub: analyze the code, correct errors, develop new functions and write documentation.
Testing showed that the model successfully solves real engineering problems, plans changes in the code, updates technology stack and performs complex commands. Claude 3.7 Sonnet has significantly improved the accuracy of work compared to previous versions.
In security matters, the model more precisely distinguishes harmful and harmless requests, reducing the percentage of false failures. Additional mechanisms of attacks against attacks through manipulation of Promtams have been developed.
Claude Code, so far available in research mode, supplements the model with the possibility of active interaction with the code, which reduces the expenses of time for routine tasks. In the nearest updates, Anthropic plans to improve the reliability of the instrument call and support long commands.