Anthropic's AI 'Claude Sonnet 4' supports 1 million token contexts, can process over 75,000 lines of code and dozens of papers at once
Anthropic's generative AI, Claude Sonnet 4, now supports context for up to 1 million tokens in the Anthropic API, five times more than before, allowing it to process entire codebases containing over 75,000 lines of code and dozens of research papers in a single request.
Claude Sonnet 4 now supports 1M tokens of context
Claude Sonnet 4's long context support is available in public beta for Anthropic API and Amazon Bedrock. It will also be available for Google Cloud's Vertex AI in the future, and the company is exploring the possibility of implementing long context in other Claude products.
By supporting longer contexts, Anthropic claims that developers will be able to use Claude to perform data-intensive use cases such as:
Large-scale code analysis
Able to read and analyze your entire code base, including source files, tests, and documentation, Claude can understand your project's architecture, identify dependencies between files, and suggest improvements that take into account the entire system design.
・Document integration
Process large document sets such as legal contracts, research papers, and technical specifications in one go, and analyze relationships between hundreds of documents while maintaining full context.
・Context-aware agents
You can also build agents that maintain context across hundreds of tool calls and multi-step workflows, incorporating full API documentation, tool definitions, and interaction history without sacrificing consistency.
With Claude Sonnet 4 now supporting contexts of up to 1 million tokens, the pricing for Claude Sonnet 4 on the Anthropic API will be adjusted only for prompts exceeding 200,000 tokens. Prompts with less than 200,000 tokens will cost $3 per million tokens for input and $15 per million tokens for output, while prompts with more than 200,000 tokens will cost $6 per million tokens for input and $22.50 per million tokens for output.
It also includes testimonials from customers who have actually used Claude Sonnet 4, which supports contexts of up to 1 million tokens.
Eric Simons, CEO of Bolt.new , which aims to transform web development by integrating Claude into its browser-based development platform, said, 'Claude Sonnet 4 is our flagship model for code generation workflows and has consistently outperformed other leading models in production environments. The adoption of 1 million context windows enables developers to tackle much larger projects while maintaining the high accuracy required for real-world coding.'
'What was once impossible has become a reality,' said Sean Ward, CEO of iGent AI , which develops Maestro, an AI partner that converts conversations into executable code. 'Claude Sonnet 4, with its 1 million token context, significantly enhances the autonomous capabilities of Maestro, iGent AI's software engineering agent. This breakthrough enables true production-scale engineering, including multi-day sessions on actual codebases, establishing a new paradigm in agent-based software engineering.'
On August 12, 2025, Anthropic added the ability to search past conversations to Claude, meaning you can continue your conversation with Claude at any time based on past conversations.
Understanding Claude's Personalization Features | Anthropic Help Center
https://support.anthropic.com/en/articles/10185728-understanding-claude-s-personalization-features
However, at the time of writing, the conversation search feature is available only to paid plan users such as Max, Team, and Enterprise, and is scheduled to be rolled out to users of other plans in the future.
A video has also been released showing how to actually use the conversation search function, which should give you an idea of how it works.
Pick up where you left off with Claude - YouTube
Related Posts:
in Software, Posted by logu_ii