Monday, the ethnic has launched a new Frontier model Is called Claud Sonnet 4.5Which it claims to provide sophisticated performance to the coding benchmarks. The agency says that Claud Sonnet 4.5 is able to create “manufactured” applications than only prototypes presenting lip in the credibility of previous AI models.
CLode Sonnet will be available through 4.5 CLOD API and CLOD Chatboat. Price for developers is Claud Sonnet 4: Equivalent to token per million input tokens (about 750,000 words, or more than the entire Lord of the ring series) and as equivalent to $ 15 per million output tokens.
Last year, anthropic AI models have emerged as a favorite of large parts of the developers and initiatives due to their strong performance in software engineering activities. Apple and Meta have been reported to have used to use Claud AI models internally and do a significant business by selling API access such as cursor, windsaroff and reflight in anthropological AI coding applications. Recently, OpenAI’s GPT -5 challenged the domination of ethnicity in space, out of clode models in various coding benches.
Anthropic says that several coding with Claud Sonnet 4.5 provides art-top performance to benchmarks, including Has been verified in the needle-benchThe However, ethnic AI researcher David Harshi told Techchen that the acting Sonnet 4.5 is difficult to capture the acting in benchmarks.
Harshi says that he has seen the Claud Sonnet 4.5 code autonomously for 30 hours during the initial examination with some enterprise customers. At that time, he not only created the AI model, but database services stand, buy the domain name and make an SOC2 audit to ensure that the product was secured.
TechCrunch Cursor’s CEO said in a statement shared with Michelle Trulyle that Claud Sonnet 4.1 represents state -of -the -art coding performances, especially in the functioning of the long horizon. Windsaroff chief executive Jeff Wang said in a statement that Claud Sonnet 4.1 represents a “new generation of coding models”.
Anthropic also claims that Claud Sonnet 4.5 is the highest aligned Frontier AI model, including low rates of psychophyse and cheating than the previous models of Claud Sonnet 4.5. The agency says it has also improved the sensitivity of Clode to launch an injection attack.
TechCrunch event
San Francisco
|
October 27-29, 2025
In addition to the introduction of Claud Sonnet 4.5, the ethnic Claud Agent is also launching SDK. The agency says it is the same infrastructure that gives the clode code strength and developers can be used to help create their own agents.
Anthropic Max is publishing a temporary study called “Imagine Claud” for customers, which shows the AI model produced software on the fly. The company says the model will respond to the user’s requests in real time without any predetermined functionality or pre -written code.