Anthropic doubles the context window to compete with GPT-5.
Anthropic now allows business clients to send much longer text prompts to its AI model Claude Sonnet 4 at once. The context window now has a million tokens, equivalent to about 750,000 words or 75,000 lines of code.
Five Times more Memory.
According to Anthropic executive Brad Abrams in TechCrunch, large context windows are particularly useful for long, complex programming tasks that can take hours.
The limit of Sonnet 4 is now five times larger, and the context window is twice as large as that of GPT-5, which offers 400,000 tokens. This makes Claude better at agentic coding tasks: the AI model can work autonomously on a problem for hours without losing important context.
Competition with GPT-5.
GPT-5 remains a strong competitor with competitive pricing and strong performance. However, Anthropic emphasizes that their model not only handles more tokens but also effectively understands them. For requests over 200,000 tokens, the API rate increases: $6 per million input tokens and $22.50 per million output tokens.
The upgrade is available via Anthropic’s API and through cloud partners such as Amazon Bedrock and Google Vertex AI. Developer platforms like GitHub Copilot, Windsurf, and Cursor make intensive use of Claude.