Anthropic is expanding the amount of information enterprise customers can send to Claude in a single prompt, a move aimed at attracting more developers to its popular AI coding models. For API customers, Claude Sonnet 4 now supports a one-million-token context window, allowing the AI to process requests as long as 750,000 words—more than the entire Lord of the Rings trilogy—or 75,000 lines of code. This marks a fivefold increase from Claude’s previous 200,000-token limit and more than doubles OpenAI’s GPT-5, which offers a 400,000-token context window.
The extended context capability will also be available for Claude Sonnet 4 through Anthropic’s cloud partners, including Amazon Bedrock and Google Cloud’s Vertex AI. Anthropic has established one of the largest enterprise businesses among AI model developers, primarily by supplying Claude to AI coding platforms like Microsoft’s GitHub Copilot, Windsurf, and Anysphere’s Cursor. While Claude remains a favorite among developers, GPT-5 poses a competitive threat with its lower pricing and strong coding performance. Notably, Anysphere CEO Michael Truell assisted OpenAI in announcing GPT-5, which is now the default AI model for new Cursor users.
Brad Abrams, Anthropic’s product lead for the Claude platform, stated in an interview that AI coding platforms will see significant benefits from this update. When questioned about GPT-5’s impact on Claude’s API usage, Abrams remained optimistic, emphasizing satisfaction with the API business’s growth. Unlike OpenAI, which relies heavily on consumer subscriptions to ChatGPT, Anthropic focuses on selling AI models to enterprises via API. This makes AI coding platforms a crucial customer base, prompting Anthropic to introduce new incentives to retain users amid GPT-5’s rise.
Recently, Anthropic released an updated version of its largest AI model, Claude Opus 4.1, further enhancing its coding capabilities.
AI models generally perform better with more context, particularly for software engineering tasks. For instance, an AI tasked with developing a new app feature will deliver better results if it can access the entire project rather than just a fragment. Abrams also highlighted that Claude’s large context window improves performance in long agentic coding tasks, where the AI autonomously works on problems for extended periods, retaining all prior steps.
However, some companies push context windows to extremes—Google’s Gemini 2.5 Pro offers 2 million tokens, while Meta’s Llama 4 Scout boasts 10 million. Studies indicate diminishing returns with excessively large context windows, as AI models struggle to process such vast inputs effectively. Abrams noted that Anthropic’s research prioritizes not just expanding the context window but also improving the “effective context window,” ensuring Claude comprehends most of the information it receives. He did not disclose the specific techniques behind this improvement.
For prompts exceeding 200,000 tokens, Anthropic will charge API users higher rates: $6 per million input tokens and $22.50 per million output tokens, up from $3 and $15, respectively.