Anthropic launches Claude Sonnet 4.5, its latest AI model
Dario Amodei, Founding Partner and General Manager of Anthropic, is at the World Economic Forum in 2025.
Stefan Wermuth | Bloomberg | Getty Images
On Monday, he announced the latest artificial intelligence model: Claude Sonnet 4.5.
The model is better in coding, use of computers and meeting practical business needs and perfect in special areas such as cyber security, finance and research. . AmazonThe backward starting start, which is worth $ 183 billion, makes Claude Sonnet 4.5 usable for all users.
Anthropic said Claude Sonnet 4.5 is the “best coding model in the world”, according to industrial criteria as a SWE-Bench, a test set that measures the software coding capabilities of the AI system.
“People are just noticing with this model, because a colleague and more of a colleague and more of a colleague, encountering problems and correcting them.” He said.
The company said that the model creates a higher quality code, it is better to determine code improvements and that it can follow the instructions more reliably.
Claude Sonnet comes after 4.5 anthropic launch Claude Opus 4.1 August and Claude Sonnet on May 4th. It is the latest example of innovation in the artificial intelligence industry.
Anthropic was founded in 2021 by a group of former Openai researchers, and the two companies have been violent competitors since then.
Openai launched a productive artificial intelligence explosion after Chatgpt Chatbot was released in 2022. Seeing that the valuation rose to $ 500 billion, the beginning announced the latest GPT-5 in August. The presentation was rocky because some users complained about the loss of access to the company’s previous models.
Mike Krieger, an anthropic’s chief product manager, said Claude Sonnet 4.5 would be assumed for users and that anthropic recommends the model for “basically every use”.
Nevertheless, users will have options. Paid subscribers can still choose to use OPUs, and users with certain workflows can choose an older generation sonnet if they are not ready to move overnight.
Krieger is smaller than Claude Sonnet 4.5’s Claude Opus 4.1, but “almost in every way”.
Antropic’s Claude Sonnet 4.5 Chatbox.
By anthropic permission
“We found this and our customers find it very useful for real, real work.” He said.
Claude Sonnet can work autonomously for 4.5 30 hours and anthropic said he could focus on complex, multi -stage tasks during that period. The Claude Opus 4, launched by the company in May, can run autonomously for only seven hours.
The company said that anthropic can improve the behavior of the model through comprehensive security training. Deception, power search, and sycophancy, such as “behaviors”, this is the time when a model tells a user what they want to hear.
Claude Sonnet 4.5 is more resistant to rapid injection attacks that a model can deceive to do something malicious, such as revealing sensitive data.
“This is the biggest security jump that we think we have seen in the last year, year, year and a half.” He said.
Additional model launch is already on the horizon for anthropic. Kaplan said that better models, including “probably opus”, came.
“No promise,” he said. “But I think we will probably have one or two versions before the end of the year.”
WRISTWATCH: Sources say Microsoft should buy AI from Openai from Openai.
