Anthropic has upgraded its Claude Sonnet 4 model to handle 1 million tokens in a single request. This allows the model to work on much larger projects, such as analyzing thousands of lines of code or multiple research papers at once.
The long-context support is currently available in public beta through Anthropic API and Amazon Bedrock and will soon be added to Google Cloud Vertex AI.
With this upgrade, developers can load entire codebases and documents in one request, making it easier to understand project structure and dependencies between files. The model can also analyze large datasets such as legal contracts, research papers, and technical documents while preserving context.
API pricing is as follows: for requests up to 200k tokens, input costs $3 per million tokens and output $15 per million tokens; for requests over 200k tokens, input costs $6 per million tokens and output $22.50 per million tokens.
The prompt caching feature reduces both costs and response times for long-context usage, and batch processing can provide up to 50% additional savings. The new features will be available to a wider user base next week.
