Anthropic launched a “Hybrid Reasoning” AI model

Anthropic launched its “Hybrid Reasoning” AI model, which can produce faster responses or show the steps for its reasoning.

This introduction is another breakthrough in AI development, with multiple tech firms, including DeepSeek, Alibaba, and OpenAI, competing against each other.

Anthropic, which Google and Amazon back, revealed that its Claude 3.7 Sonnet model is its most advanced.

While offered for all users, its extended thinking mode is only available on paid plans.

Claude 3.7 Sonnet does self-reflection before responding to questions. The company said this improves its ability to answer math, physics, and coding problems, among other things.

It added that the Sonnet was designed to focus less on computer science and math and more on real-world tasks, reflecting how businesses use large language models (LLMs).

The startup also revealed it’s releasing Claude Code (limited-release preview), an AI agentic code tool that aims to assist developers with coding.

Screenshot from Anthropic

The AI-powered agentic application allows developers to perform coding-related tasks autonomously.

What makes it different?

“We fundamentally believe that reasoning is a feature of the AI rather than a completely separate thing,” Dianne Penn, product research lead at Anthropic, told the Verge, to simplify the experience of using a model unlike OpenAI and other companies offering reasoning models as separate products.

Where Is Claude 3.7 Sonnet Available?

It’s available in the Claude app and through Anthropic’s API, Google Cloud’s Vertex AI, and Amazon Bedrock (for developers).

And the cost? It’s the same as 3.5 Sonnet, its predecessor—$3 per million input tokens and $15 per million output tokens for paid users.

Users can select how much time and resources the AI will devote to answer a question, allowing them to cap the thinking involved by telling Claude 3.7 not to spend over a specific amount of tokens.