Claude Sonnet 4 supports up to 1 million talk contexts with human APIs. This allows you to process your entire codebase in one request with over 75,000 lines of code or dozens of research papers.
Sonnet 4's long context support is found in the public beta of human APIs and Amazon Bedrock, with Google Cloud's pinnacle AI coming soon.
Longer contexts, more use cases
As the context grows, developers can perform more comprehensive, data-intensive use cases, including:
- Large-scale code analysis: Loads the entire codebase, including source files, tests, and documents. Claude can understand the project architecture, identify cross-file dependencies, and propose improvements that explain the complete system design.
- Document Integration: Handles a wide set of documents including legal contracts, research papers, and technical specifications. Analyze relationships between hundreds of documents while maintaining full context.
- Context Aware Agent: Build agents that maintain context across hundreds of tool invocations and multi-step workflows. Include the complete API documentation, tool definitions, and interaction history without inconsistent.
API Pricing
To account for increased calculation requirements, pricing adjusts prompts above 200k tokens.
| Prompt ≤200k | Prompt> 200k | |
|---|---|---|
| input | $3/mtok | $6/mtok |
| output | $15 / mtok | $22.50/mtok |
When combined with quick caching, users can reduce latency and costs of Claude Sonnet 4 in a long context. The 1M context window can also be used with a 50% cost savings in batch processing.
Customer Spotlight: Bolt.New
Bolt.New transforms web development by integrating Claude into a browser-based development platform.
“Claude Sonnet 4 remains the go-to model for code generation workflows, consistently outperforming other major models in production. With a 1M context window, developers can tackle fairly large projects while maintaining the high accuracy required for real-world coding.”
Customer spotlight: opgent ai
Based in London, AI is working in the field of software development at Maestro, an AI partner that transforms conversations into executable code.
“The reality is that it was once impossible. With a 1M token context, Claude Sonnet 4 is supercharged with the autonomous capabilities of Maestro, the software engineering agent for Eigent AI.
Let's get started
Sonnet 4's long context support is a public beta of the Human API for customers with Tier 4 and custom rate limits, with wider availability rolling out over the coming weeks. Amazon Bedrock also offers long-term contexts and will soon be available in Google Cloud's Vertex AI. They are also looking for ways to bring a long-term context to other Claude products.
For more information about Sonnet 4 and the 1M context window, see the documentation and pricing page.
