(Credit: Anthropic)
Anthropic released a new AI reasoning model today called Claude 3.7 Sonnet, as well as an agentic AI coding tool called Claude Code.
Claude 3.7 Sonnet can manage two types of information processing at once, which is why Anthropic is calling it the "first hybrid reasoning model." It can either produce "near-instant responses" or engage in a drawn out, step-by-step train of thought that's visible to the user.
Claude 3.7 Sonnet is now available on all plans—including Free, Pro, Team, and Enterprise—as well as the Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI. But the extended thinking mode is not available on the free tier.
AI models that walk users through how they arrive at their answers is a big trend in chatbot land, and considered a key technological advancement. OpenAI released its version, GPT-o1, in September 2024. DeepSeek also has a "DeepThink" mode, as does Elon Musk's Grok 3 chatbot, which is how a user caught it censoring negative information about Musk and Trump.
Claude 3.7 Sonnet differentiates itself from DeepSeek and OpenAI by combining the two thinking types in one: "Just as humans use a single brain for both quick responses and deep reflection, we believe reasoning should be an integrated capability of frontier models rather than a separate model entirely," Anthropic says.
Users can pick if they want the model to answer quickly, or to think longer before answering. In extended thinking mode, the model "self-reflects before answering, which improves its performance on math, physics, instruction-following, coding, and many other tasks," Anthropic says.
This type of unified model experience, where users don't need to manually select different models for different tasks, is another AI chatbot trend. Earlier this month, OpenAI CEO Sam Altman said the company will eventually remove ChatGPT's "model picker."
"We want AI to 'just work' for you; we realize how complicated our model and product offerings have gotten," Altman says. "We hate the model picker as much as you do and want to return to magic unified intelligence."
Developers who access Claude 3.7 Sonnet through Anthropic's API can also set a "budget" for how much computational effort the model uses to answer. "This allows you to trade off speed (and cost) for quality of answer," Anthropic says. The company made a point to say that Claude 3.7 Sonnet "has the same price as its predecessors," likely in an attempt to compete with DeepSeek, which grabbed headlines for offering high intelligence without an increase in computing cost.
Finally, Claude's latest release hits on perhaps the biggest trend in AI right now—agentic, or autonomous, functionality—with Claude Code. It's only available in a limited research preview for now, but promises to revolutionize coding by doing part of a developers' job for them.
Claude is known for its strong coding chops, and with Claude Code developers can "delegate substantial engineering tasks to Claude directly from their terminal." It acts 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—keeping you in the loop at every step."
Anthropic says these types of agentic capabilities are a step above the tech it debuted in 2024, and on the road to a world where Claude can "find breakthrough solutions" on its own in 2027.
