On Monday, humanity launched a new frontier model called Claude Sonnet 4.5. It claims to provide cutting-edge performance in code benchmarks. The company says that Claude Sonnet 4.5 not only represents a leap in reliability from previous AI models, but also allows you to build “production-ready” applications rather than just a prototype.
Claude Sonnet 4.5 is available with the Claude API and Claude Chatbot. Developers’ prices are the same as Claude Sonnet 4: $3 per million (approximately 750,000 words, or the entire “Lord of the Rings” series) and $15 per $100 output token.
Last year, human AI models emerged as a favorite among developers and businesses due to the strong performance of software engineering tasks. Apple and Meta reportedly use the Claude AI model internally, and Anthropic is in the key business of selling API access to AI coding applications such as Cursor, Windsurf and Replit. Recently, Openai’s GPT-5 has challenged humanity’s domination in space, surpassing the Claude model with various coding benchmarks.
Humanity says that Claude Sonnet 4.5 offers industry-leading performance in several coding benchmarks, including SWE bench validation. However, human AI researcher David Hershey tells TechCrunch that capturing the performance of the Claude Sonnet 4.5 with Benchmarks alone is difficult.
Hershey says he viewed the Claude Sonnet 4.5 code autonomously for up to 30 hours during early testing with clients from some companies. Meanwhile, he saw the AI model not only build applications, but also stand up database services, buy domain names, and run SOC 2 audits to ensure the product is secure.
In a statement shared with TechCrunch, Cursor CEO Michael Truell said Claude Sonnet 4.5 represents cutting-edge coding performance, especially for long tasks on Horizon tasks. Windsurf CEO Jeff Wang said in a statement that Claude Sonnet 4.5 represents a “new generation coding model.”
Humanity also claims that Claude Sonnet 4.5 is the most consistent frontier AI model to date, with a lower rate of psychofancy and deception than previous models. The company says Claude’s sensitivity to rapid injection attacks has also improved.
TechCrunch Events
San Francisco
|
October 27th-29th, 2025
In addition to the release of Claude Sonnet 4.5, Anthropic also releases the Claude Agent SDK. The company says this is the same as the infrastructure that enhances Claude Code, and can be used to allow developers to build their own agents.
Humanity is releasing a temporary research preview called “Imagine With Claude,” a MAX subscriber showing AI model generation software on the fly. The company says the model responds to user requests in real time without pre-determined features or pre-written code.
Tensional competition in the AI world has made it common for businesses to ship flagship models every few months. The Claude Sonnet 4.5 will be released within two months of Anthropic’s last AI model, the Claude Opus 4.1. These rapid production cycles make it difficult for any company to hold meaningful leads for a very long time.
Source link
