Anthropic Unveils Claude 3.7 Sonnet and Claude Code for Advanced AI Reasoning

Anthropic has unveiled Claude 3.7 Sonnet, its most advanced AI model yet and the first hybrid reasoning model available. Designed to combine rapid response generation with deep, structured problem-solving, Claude 3.7 Sonnet offers an extended thinking mode. Through the API, users can fine-tune the model’s reasoning time, balancing speed, cost, and accuracy for greater control over its depth of analysis.

Claude 3.7 Sonnet brings significant enhancements to coding and front-end web development. Alongside this model, Anthropic has introduced Claude Code, an agentic coding tool designed to assist developers with complex engineering tasks directly from their terminal.

Claude 3.7 Sonnet is now available across all Claude plans—including Free, Pro, Team, and Enterprise—as well as through the Anthropic API, Amazon Bedrock, and Google Cloud’s Vertex AI. However, its extended thinking mode is not available on the free tier. Pricing remains consistent with previous versions: $3 per million input tokens and $15 per million output tokens, which includes tokens used for extended reasoning.

A New Approach to AI Reasoning

Anthropic designed Claude 3.7 Sonnet with an integrated reasoning approach, moving away from traditional AI models that separate rapid response generation from complex problem-solving. Just as humans seamlessly switch between quick thinking and deep reflection, we believe advanced AI models should integrate both capabilities rather than rely on separate systems for reasoning.

• Dual-Mode Reasoning: Claude 3.7 Sonnet functions as both a standard LLM (large language model) and a deep reasoning system, allowing users to switch between quick responses and extended reflection. The latter improves performance in math, physics, coding, and instruction-following tasks.

• API-Controlled Thinking Budget: Developers using the Claude API can set a maximum number of tokens for reasoning, for any value up to the model’s 128K token output limit, providing flexibility in speed, cost, and answer quality.

• Optimized for Real-World Applications: Unlike models tuned primarily for academic benchmarks, Claude 3.7 Sonnet is designed for practical business applications, making it more effective in real-world problem-solving scenarios.

Early evaluations highlight its best-in-class coding capabilities, outperforming competitors in handling complex codebases, tool integration, and full-stack development. Companies such as Cursor, Cognition, Vercel, and Replit report that Claude 3.7 Sonnet excels in generating production-ready code with superior precision and fewer errors. Canva's testing found it significantly reduced development time while maintaining high design quality.

Introducing Claude Code: A Developer’s AI Assistant

Anthropic has also launched Claude Code, an agentic coding tool designed to enhance software development workflows. Available as a limited research preview, Claude Code enables developers to:

• Search, read, and edit code directly from their command line.

• Write and run tests, commit changes, and push code to GitHub.

• Interact with command-line tools, maintaining developer oversight at every step.

Internal testing shows that Claude Code automates coding tasks that typically take 45+ minutes to complete manually, streamlining debugging, test-driven development, and large-scale code refactoring, saving development tie and overhead. Anthropic plans to refine Claude Code in the coming weeks by improving tool reliability, long-running command support, and in-app rendering.

The Claude GitHub integration has also been expanded across all Claude plans, allowing users to seamlessly connect their repositories to debug, develop, and document code with AI assistance.

Commitment to Responsible AI Development

Anthropic has subjected Claude 3.7 Sonnet to extensive testing to ensure its safety, reliability, and ethical AI compliance. Compared to previous models, it reduces unnecessary refusals by 45%, making it more responsive while maintaining security measures.

The Claude 3.7 Sonnet system card outlines safety evaluations, including defenses against prompt injection attacks and assessments of AI reasoning transparency. Researchers can use these insights to improve AI safety practices.

Looking Ahead

The launch of Claude 3.7 Sonnet and Claude Code represents a significant step toward AI-powered problem-solving and autonomous development. As AI models become more capable of deep reasoning and collaborative coding, they are poised to enhance human productivity and creativity in unprecedented ways.

Anthropic continues to refine its AI models, prioritizing practical applications, developer tools, and responsible AI scaling. With future updates on the horizon, Claude’s evolution will further expand the possibilities of intelligent automation and human-AI collaboration. Meanwhile, as AI technology advances, other companies—including OpenAI with its upcoming GPT-5—are also developing hybrid models that integrate rapid responses with deeper reasoning, reflecting a broader industry shift toward more adaptable AI systems. This competition among AI leaders is driving innovation, shaping the next generation of intelligent assistants, and redefining how humans and machines work together.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.