Another week and another new AI model is ready for public use. This time, Anthropic has introduced Claude 3.7 Sonnet. The company describes its latest release as the first “hybrid thinking model” on the market, which means that the new version of Claude can either answer a question almost instantly or take its time, working through it step by step. You, as the user, can decide which approach Claude takes with a drop-down menu that allows you to select the “thinking mode” you want it to use.
“We designed Claude 3.7 Sonnet with a philosophy that is different from other thinking models on the market. Just as humans use one brain for both quick responses and deep thinking, we believe that reasoning should be an integrated capability of edge models, not a separate model,” Anthropic writes. “This unified approach also creates a more seamless experience for users.”
Anthropic doesn’t name OpenAI directly, but the company is clearly taking aim at its competitor. Between the GPT-4, the o1, the o1-mini, and now the o3-mini, OpenAI offers many different models, but if you don’t follow the company closely, the number of systems on offer can be overwhelming; in fact, Sam Altman recently admitted as much. “We hate model selection as much as you do and want to get back to the magic of unified intelligence,” he wrote on X earlier this month.
Anthropic says they also took a different approach to developing Claude’s reasoning. “We have optimized it somewhat less for math and computer science olympiad problems, instead shifting the focus to real-world problems that better reflect how businesses actually use masters,” the company writes. Thus, current Claude users can expect “particularly strong improvements in coding and front-end web development.”
Claude 3.7 Sonnet is available for use starting today for all Claude plans, including the free Anthropic plan. Meanwhile, developers can access the new model through the company’s API, Amazon Bedrock, and Vertex AI on Google Cloud.
Speaking of developers, Anthropic is also introducing Claude Code, a new “agent” tool that allows delegating coding tasks to Claude directly from the terminal interface. Currently available as a limited trial version, Anthropic says Claude Code can read code, edit files, write and run tests, and even submit commits to GitHub.
