Anthropic AI releases Claude 3.7 Sonnet, their most AI advanced model yet

Anthropic AI has unveiled its latest AI model, the Claude 3.7 Sonnet, their first hybrid reasoning model. Built on the foundation of the Claude 3.6 Sonnet and Claude 3.5 Sonnet, the model is able to solve more complex queries while also improving in subjects like math and coding. 

Besides, the AI startup also released a ‘limited research preview’ of Claude Code, a coding tool. Anthropic described the assistant as “an active collaborator that can search and read code, edit files, write and run tests, commit and push code to GitHub, and use command line tools,” in the blog making the announcement. 

Just like other AI reasoning models, Claude 3.7 Sonnet has chain of thought and an extended thinking mode which allows users to set for how long the model should spend on a question in terms of tokens. 

The model has scored 62.3 percent on the SWE-Bench which tests GitHub issue-solving abilities, and 70.3 percent with extended thinking mode on, which is higher than OpenAI’s GPT-4o. On the Multi-Lingual Understanding benchmark, it scored 83.2 percent and 86 percent with extended thinking. The Claude 3.7 Sonnet also surpassed GPT-4o at agentic and coding tasks. 

Starting Monday, Claude 3.7 Sonnet will be available in the Claude app through Anthropic’s API, Amazon Bedrock and Google Cloud’s Vertex AI at a rate of $3 per million input tokens and $15 per million output tokens which is higher than GPT-40 Mini available at $1.10 per million input tokens and $4.40 per million output tokens.

Separately, a former Google Brain researcher and one of the authors of the seminal paper on Transformers that kickstarted the shift towards GenAI technology, Niki Parmar has joined Anthropic AI. Parmar, who quit Google in 2021 to co-found Adept AI Labs and Essential AI announced the news on X saying she was “thrilled to have contributed to the development of the Claude 3.7,” after joining the OpenAI rival in December last year.