Anthropic announced the expansion of its AI model Claude Sonnet 4, which can now process up to 1 million tokens in a single request via the API. This allows developers and companies to analyze large volumes of text or entire software projects without the need to split them into smaller parts. The new context volume corresponds to approximately 750,000 words or over 75,000 lines of code, enabling the model to work simultaneously with complete software product architectures or hundreds of documents.
The feature is already available for Anthropic API users with high limits, as well as on Amazon Bedrock platforms and will soon be available on Google Cloud Vertex AI. Company representative Brad Abrams noted that the context expansion is particularly useful for platforms that use AI for programming, as Claude can now perform long, multi-step tasks while maintaining logic and relationships between different parts of the code.
The company has changed the pricing: for requests over 200,000 tokens, the cost is $6 per million input tokens and $22.50 per million output tokens. This allows covering the additional resources required for processing large volumes of data. Clients from various industries, including software development, finance, and legal services, have already started using the new capability to analyze large datasets and complex projects.
Claude Sonnet 4 demonstrates high accuracy when working with large contexts, particularly in searching for specific information within large volumes of text. This expansion helps development and analytics teams work on projects without losing important connections between parts of the information. Anthropic plans to gradually provide access to the new feature to a wider range of users in the coming weeks.
