Claude 3.7 Sonnet just dropped, proving once again that the AI arms race is less about world domination and more about which chatbot can be your smartest, least annoying coworker.

Claude 3.7 Sonnet is its first-ever “hybrid reasoning model,” designed to tackle more complex challenges and outperform its predecessors in areas like math and coding.

Alongside the new model, the company is introducing a “limited research preview” of Claude Code, an AI-powered coding assistant built to handle more hands-on development work.

While competitors like OpenAI offer separate models specialized in reasoning, Anthropic is betting on a more streamlined experience—one AI that does it all.

Claude 3.7 Sonnet shows noticeable improvements in agentic coding, finance, and legal tasks, making it a step up for users needing AI-powered precision.

Anthropic, which already supports AI coding tools like Cursor, is pitching Claude Code as a true collaborator—it can search and analyze code, edit files, write and run tests, commit and push to GitHub, and even handle command-line tools.

The new model is available starting Monday in the Claude app, with API access through Anthropic, Amazon Bedrock, and Google Cloud’s Vertex AI.

Pricing remains the same as its predecessor, 3.5 Sonnet, at $3 per million input tokens and $15 per million output tokens.

Anthropic employees have already used Claude 3.7 Sonnet to build website front-ends, interactive games, and even spend up to 45 minutes refining test cases in an iterative coding process, according to Anthropic’s Penn.

While Claude still doesn’t have real-time web search like some competitors, its October 2024 knowledge cut-off makes it one of the most up-to-date AI models available.

The AI race is moving fast—just ask Elon Musk, who unveiled Grok-3 last week. And Perplexity just launched a web browser today.

But with Claude 3.7 Sonnet’s performance gains, Anthropic looks to be pulling ahead again. And if this release is any indication, the industry’s future might not be about separate reasoning models, but one AI that can handle it all.