Claude Sonnet 4 now supports a 1M-token context window

Anthropic is dramatically expanding the capabilities of its AI coding models, announcing that enterprise customers can now send up to one million tokens in a single prompt to its Claude Sonnet 4 model. That’s roughly 750,000 words — enough to fit The Lord of the Rings trilogy with room to spare — or about 75,000 lines of code.

The update represents a fivefold jump from Claude’s previous 200,000-token limit, and it more than doubles the 400,000-token context window offered by OpenAI’s GPT-5. Long-context support will be available not just via Anthropic’s own API, but also through cloud partners including Amazon Bedrock and Google Cloud’s Vertex AI.

A Strategic Move Against GPT-5

Anthropic has built a substantial enterprise customer base, particularly among AI coding platforms such as Microsoft’s GitHub Copilot, Windsurf, and Anysphere’s Cursor. While Claude is the preferred choice for many developers, GPT-5’s competitive pricing and strong coding performance pose a potential challenge. Anysphere CEO Michael Truell, notably, helped announce GPT-5, which is now Cursor’s default AI model for new users.

Brad Abrams, Anthropic’s product lead for Claude, downplayed concerns about competition in an interview with TechCrunch, saying he’s “really happy with the API business and the way it’s been growing.” Abrams also suggested the update will offer “a lot of benefit” to AI coding platforms, particularly for long, autonomous coding tasks where the model needs to retain awareness of previous steps over extended sessions.

Why Context Matters in Coding

For software engineering, larger context windows can be a game changer. When an AI can see the full scope of a project — not just isolated snippets — it can generate more coherent, integrated features. This is especially critical for “agentic” coding tasks, where the AI is essentially working on its own for minutes or hours.

While Anthropic’s upgrade is substantial, it’s not the largest in the industry. Google’s Gemini 2.5 Pro offers a 2 million token context window, and Meta’s Llama 4 Scout boasts a staggering 10 million tokens. However, Abrams noted that Anthropic has focused on boosting the “effective context window” — the portion of the prompt the AI can genuinely comprehend — rather than just inflating numbers. The company has not disclosed exactly how it achieves this.

Pricing and Availability

The expanded context window comes with higher costs for large requests. For prompts over 200,000 tokens, API customers will pay $6 per million input tokens and $22.50 per million output tokens, up from $3 and $15 respectively.

This move follows last week’s release of Claude Opus 4.1, Anthropic’s most capable model to date, which improved coding abilities further. It underscores the company’s enterprise-first strategy — contrasting with OpenAI’s consumer-focused approach — and signals that the race for developer mindshare in AI coding tools is only intensifying.