On Monday, Anthropic launched a new frontier model Claude Sonnet 4.5Which claims to offer state -of -the -art performance on the coding benchmark. The company says that Cloud Sonnet 4.5 is capable of manufacturing “production-taiyar” applications rather than prototype, which represents a jump in credibility from the previous AI model.
Cloud Sonnet 4.5 will be available in Cloud API and Cloud Chatbot. Pricing Cloud Sonnet for Developers: $ 3 per million is similar to input tokens (about 750,000 words, or more than the entire Lord of the rings series) and $ 15 per million per million per million output tokens.
In the previous year, AI models of Anthropic have emerged as a favorite among developers and enterprises, in large parts due to their strong performance on software engineering works. Apple and META allegedly use cloud AI models internally, and Anthropic has created an important business selling API access for AI coding applications such as cursor, windsurf, and reply. Recently, Openai’s GPT-5 has challenged the dominance of the anthropic in space, improving the cloud model on a variety of coding benchmarks.
Anthropic states that Cloud Sonnet 4.5 provides industry-agar performance on several coding benchmarks, including Self-bench verifiedHowever, an anthropic AI researcher David Hershe told Techchrech that it is difficult to capture the performance of Cloud Sonnet 4.5 on the benchmark alone.
Hershe says that he has observed the Cloud Sonnet 4.5 code for 30 hours during initial tests with some enterprise customers. At that time, he not only created an application to the AI model, but also erected database services, buy domain names, and to ensure that the product was safe to ensure that the product was safe.
In a statement shared with the CEO of Techcrunch Cursor, Mitchell Truel said that Cloud Sonnet represents 4.5 state -of -the -art coding performance, especially on long -term horizon functions. Windsurf CEO Jeff Wang said in a statement that Cloud Sonnet represents 4.5 “new generation of coding models”.
Anthropic also claims that Cloud Sonnet 4.5 is its most alignment Frontier AI model, compared to previous models with lower rates of sycophancy and deception. The company says that it has also improved the sensitivity of the cloud to accelerate the injection attacks.
Techcrunch event
San francisco
,
27-29 October, 2025
Along with the launch of Cloud Sonnet 4.5 -Anthropic is also launching Cloud Agent SDK. The company says that this is the same infrastructure that powers the cloud code, and it can be used to help developers in the manufacture of their own agents.
Anthropic Max is also issuing a temporary research preview called “Imagine with Cloud” for subscribers, which shows the AI model generating software on the fly. The company says that the model will respond to the user’s requests in real time, with no predetermined functionality or predetermined code.
Stressful competition in the AI world has made it common for companies to ship the flagship model every few months. Cloud Sonnet 4.5 is starting for less than two months of Anthropic’s final AI model, Cloud Opus 4.1. These rapid production cycles make it difficult for any company to gain meaningful lead for a very long time.
