Anthropic has announced the launch of its new AI model, Claude Sonnet 4.5, which it claims is the most productive in the programming world. According to the developers, the model is capable of creating not only prototypes, but also "production-ready" applications, which is a significant step forward in code reliability compared to previous AI models.
Claude Sonnet 4.5 is available via the API and the Claude chatbot. The cost for developers remains the same as the previous version of Sonnet 4: $3 per million incoming tokens and $15 per million outgoing tokens.
As TechCrunch writes, over the past year, Anthropic models have been actively used in the field of software development. Apple and Meta are known to use them in their internal processes. In addition, the Claude API is integrated into the Cursor, Windsurf, and Replit services. At the same time, competition in the market has increased after the release of GPT-5 from OpenAI, which surpassed previous Anthropic models in a number of tests.
The company claims that Claude Sonnet 4.5 performs best in the industry in several benchmarks, including SWE-Bench Verified. Anthropic researcher David Hershey reported that during testing, the model was able to run autonomously for more than 30 hours, building an application, configuring databases, purchasing domain names, and even conducting a SOC 2 audit to verify the product's security.
Anthropic emphasizes that the model has become more consistent in its responses, and that it has reduced the level of flattery and false claims. In addition, it has improved protection against attacks through "rapid injections."
Along with the launch of Claude Sonnet 4.5, the company introduced the Claude Agent SDK, a set of tools for building custom agents based on the same infrastructure used in Claude Code. In addition, Max subscribers have a preview of the "Imagine with Claude" feature, which demonstrates real-time software generation.
The release of Claude Sonnet 4.5 took place less than two months after the release of Claude Opus 4.1, which indicates the high pace of model updates in the field of artificial intelligence.