Loading post…
AnthropicAnthropicAnnouncement
Claude 3.7 Sonnet and Claude Code
Summary
Claude 3.7 Sonnet is Anthropic’s most intelligent model and the first hybrid reasoning model, combining near-instant responses with extended step-by-step thinking visible to users for improved math, science, and coding tasks. API users have fine-grained control over the thinking budget, allowing them to specify up to 128K tokens for deeper reasoning. Available on all Claude plans, Amazon Bedrock, and Vertex AI, the model achieves state-of-the-art performance on SWE-bench Verified and TAU-bench, demonstrating strong improvements in coding and front-end web development. Anthropic also introduced Claude Code, a command line tool for agentic coding in limited research preview, capable of searching and editing files, running tests, and committing code to GitHub. Pricing remains at $3 per million input tokens and $15 per million output tokens including thinking tokens, while the model reduces unnecessary refusals by 45% compared to its predecessor.
Context
Other reasoning models on the market are often separate from standard LLMs, but Claude 3.7 Sonnet integrates reasoning as a unified capability, allowing users to seamlessly switch between normal and extended thinking modes.
Approach / What changed
Claude 3.7 Sonnet is a hybrid reasoning model that combines an ordinary LLM with a reasoning model in one, offering both standard and extended thinking modes with fine-grained API control over the thinking budget up to 128K tokens. Claude Code is a CLI tool for agentic coding with capabilities such as file editing, command execution, and GitHub integration, currently available in limited research preview.
Takeaways
Claude 3.7 Sonnet is a hybrid reasoning model that can produce instant responses or extended step-by-step thinking, with API users able to control the thinking budget up to 128K tokens.
Claude Code is a command line tool for agentic coding that can search and read code, edit files, run tests, and commit to GitHub, available in limited research preview.
Pricing for Claude 3.7 Sonnet remains at $3 per million input tokens and $15 per million output tokens including thinking tokens, while unnecessary refusals are reduced by 45% compared to its predecessor.
FAQ
What improvements does Claude 3.7 Sonnet show in coding?
Claude 3.7 Sonnet shows strong improvements in coding and front-end web development, achieving state-of-the-art performance on SWE-bench Verified and TAU-bench. Early testing from Cursor, Cognition, Vercel, Replit, and Canva reported it as best-in-class for real-world coding tasks, handling complex codebases and full-stack updates with reduced errors.
How can developers control the thinking budget in the API?
API users can set a maximum thinking budget up to 128K tokens, allowing them to trade off speed and cost for quality of answer. The model will think for no more than the specified number of tokens before responding.
What is Claude Code and what can it do?
Claude Code is a command line tool for agentic coding available in limited research preview. It can search and read code, edit files, write and run tests, commit and push code to GitHub, and use command line tools, keeping developers in the loop at every step.
Source and freshness
- Source
- anthropic.com
- Published
- Added to Yomu
Summaries are generated from the original by a language model. Check the original before relying on it.
Keep reading
Related reading
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
The best engineering reads, once a week
One useful email. Unsubscribe anytime.