GPT-5 and Anthropic have updated the Claude Sonnet 4 model, expanding the context window to 1 million tokens. This increase allows the model to process approximately 750,000 words or 75,000 lines of code simultaneously. This new limit doubles GPT-5’s 400,000-token capacity. This will allow Claude to support larger and longer-term tasks, particularly in enterprise applications.
Claude surpasses GPT-5 in this area
Limited to 200,000 tokens in the previous version, Claude can now read and analyze five times more data at once. This expansion is a critical improvement for the model’s ability to follow a task from beginning to end. It is also being evaluated for its ability to maintain consistency by remembering previous steps. Anthropic product leader Brad Abrams stated that this capacity will be particularly beneficial in long, multi-step software development processes.
The new context support is not limited to Anthropic’s own platform. Claude is also accessible via Sonnet 4, Amazon Bedrock, and Google Cloud Vertex AI. This broad access allows companies using different infrastructures to easily integrate the model.
Claude’s context window of 1 million tokens is one of the highest in the AI field, but it’s not an absolute record. Google reached 2 million tokens with its Gemini 2.5 Pro model. Additionally, Meta reached the 10 million token limit with its Llama 4 Scout model.
However, just as important as the size of the context window is how effectively the model uses this information. Anthropic states that Claude can not only read data but also actively evaluate a large portion of the context. This feature is described as the model’s “active context window.”
Pricing has also changed with the expanded context. API fees increase for usage exceeding 200,000 tokens. The fee for 1 million input tokens is $6, and for an output token, it is $22.50. This represents an increase of approximately two times the previous pricing. Previously, the same operations cost $3 and $15, respectively.
Claude’s increased capabilities have allowed it to gain a strong position, particularly on software development-focused platforms. Tools like GitHub Copilot, Windsurf, and Anysphere Cursor prefer Claude models for their AI integrations.
