Anthropic has unveiled a powerful expansion to its AI model Claude Sonnet 4, increasing its context window from 200,000 tokens to a massive 1 million tokens—a fivefold enhancement that empowers the model to comprehend and process extraordinary volumes of text and code.
What This Means for Users
- Scale-rich comprehension: Claude Sonnet 4 can now analyze up to 2,500 pages of text, entire codebases with 75,000–110,000 lines, or dozens of research documents in a single request.
- Reduced fragmentation: Developers no longer need to break projects into smaller chunks. As Anthropic’s product lead Brad Abrams explained: “They have to break up their problems into these small chunks… with a million tokens, the model can handle the entire scope…” The Verge
Availability & Pricing
- The 1 million-token window is live in public beta via the Anthropic API for high-tier (Tier 4) or custom-rate-limit customers. It’s also accessible through Amazon Bedrock, and coming soon to Google Cloud’s Vertex AI. More comprehensive availability is expected in the coming weeks.
- Pricing for token usage is tiered:
- Up to 200K tokens: $3 per million input tokens, $15 per million output tokens
- Above 200K tokens: $6 per million input tokens, $22.50 per million output tokens
- Users can leverage prompt caching and batch processing to reduce latency and costs by up to 50%.
Real-World Applications
- Bolt.new, which uses Claude for browser-based development, reports being able to tackle larger projects with improved accuracy thanks to the expanded context.
- iGent AI’s Maestro—an AI software engineering assistant—now supports multi-day, real-world codebase sessions, enabled by the new context capacity.
- VentureBeat notes the upgrade lets Claude “process entire software projects or dozens of research papers without breaking them into smaller chunks.”
Competitive Context
This move positions Anthropic closer to parity with OpenAI’s GPT-4.1, which also recently offered a 1 million-token context window. Anthropic’s focus on accuracy and effective context processing, not just raw capacity, is its strategic advantage.
