Key takeaways
- Expand context window for Claude Sonnet 4 to 1 million tokens in public beta.
- Enable comprehensive workflows in large-scale code analysis and document processing.
- Adjust pricing for prompts over 200,000 tokens to reflect higher compute costs.
- Introduce prompt caching and batch processing to reduce latency and costs.
- Plan broader rollout and integration of long context into other Claude products.
Anthropic has rolled out extended context support for Claude Sonnet 4, expanding its capacity to handle up to 1 million tokens on the Anthropic API. The feature, now in public beta for Tier 4 (service levels) and custom rate-limit customers, is also available on Amazon Bedrock, with Google Cloud’s Vertex AI support coming soon.
The update represents a fivefold increase from the previous limit, enabling use cases such as analysing codebases exceeding 75,000 lines, synthesising dozens of research papers, and powering agents that maintain context across hundreds of tool calls. Anthropic says this unlocks more comprehensive workflows in large-scale code analysis, document processing, and multi-step automation.
Pricing for prompts exceeding 200,000 tokens has been adjusted to reflect higher compute costs, at $6 per million tokens for input and $22.50 per million tokens for output. Meanwhile, prompts within 200K remain priced at $3 and $15, respectively.
Prompt caching and batch processing are available to cut latency and costs, with batch mode offering up to 50% savings. Early adopters are already testing the expanded window.
“With the 1M context window, developers can now work on significantly larger projects while maintaining the high accuracy we need for real-world coding,” said Eric Simons, CEO of Bolt.new.
Sean Ward, CEO and Co-founder of iGent AI, said, “This leap unlocks true production-scale engineering—multi-day sessions on real-world codebases—establishing a new paradigm in agentic software engineering.”
Anthropic plans a broader rollout in the coming weeks and is exploring the integration of long context into other Claude products. With the new context window limit, Anthropic might bounce back to compete with Google Gemini for users looking for the same.
