Anthropic has quintupled the prompt capacity of its Claude Sonnet 4 AI model for enterprise API customers, allowing users to send up to 1 million tokens in a single request, equivalent to roughly 750,000 words, 75,000 lines of code, or five times its previous limit of 200,000 tokens.

The move surpasses OpenAI’s GPT-5, which offers a 400,000-token limit, and positions Claude as a leader in handling large-scale AI coding tasks.

The expanded context window is available via Anthropic’s own API as well as cloud platforms such as Amazon Bedrock and Google Cloud’s Vertex AI. Anthropic, known for its strong enterprise base, has built a significant share of the AI coding market through integrations with Microsoft’s GitHub Copilot, Windsurf, and Anysphere’s Cursor. However, competition from GPT-5 is intensifying, with some platforms switching to OpenAI’s model.

Brad Abrams, Anthropic’s product lead, said the update will particularly benefit AI coding platforms by enabling more complex, long-horizon development tasks. He noted that the company focused not only on increasing the “context window” but also on improving its “effective” processing ability — ensuring the model can meaningfully use the information provided.

Large context windows are especially valuable for software engineering, where seeing an entire codebase can improve feature development and debugging. They also enhance “agentic” tasks, where AI works autonomously for extended periods.

Anthropic’s pricing will rise for prompts exceeding 200,000 tokens, to $6 per million input tokens and \$22.50 per million output tokens. The upgrade follows last week’s release of Claude Opus 4.1, the company’s most capable AI model to date.

While Google’s Gemini 2.5 Pro offers a 2 million-token limit and Meta’s Llama 4 Scout boasts 10 million, industry experts caution that raw size doesn’t always translate to better results.