Anthropic pushes Claude to a 1M-token context to rival OpenAI

Anthropic has raised the context window for Claude Sonnet 4 to 1 million tokens, a fivefold jump that aims squarely at the AI coding market. A larger context window is the model’s working memory, letting it consider far more text at once.  Anthropic says the upgrade allows the model to read about 2,500 pages in one go—enough to fit a full copy of War and Peace—and to analyze dozens of research papers or hundreds of documents in a single API request, roughly doubling what a 500,000-token window could handle.  For coding, the gains are sharper: moving from about 20,000 lines of code at a 200,000-token window to entire codebases of roughly 75,000 to 110,000 lines. Brad Abrams, the product lead for Claude, said customers have struggled to split projects into smaller pieces; with a million tokens, the model can take in the full scope and solve problems at their actual scale. The move keeps pace with rivals.  OpenAI offered a similar 1 million-token context in April with GPT-4.1 and, just last week, launched GPT-5, highlighting stronger coding results. Enterprise buyers are spending for reliable coding help, giving both companies a clear path to revenue while they invest heavily in infrastructure.  Anthropic has been shipping rapidly: Opus 4 and Sonnet 4 arrived about two and a half months ago, Opus 4.1 landed a week ago, and now the 1 million-token context follows. The company is reportedly seeking new funding that could value it near $170 billion, and stronger coding performance supports that pitch.  Early access to the expanded window starts today through the Anthropic API for customers on Tier 4 and custom rate limits—generally heavier users—with wider rollout planned in the coming weeks. Interest is strongest in software, pharmaceuticals, retail, professional services, and legal. The upgrade reduces the need to break up files or repositories and reinforces Anthropic’s bid to reclaim momentum in the ongoing AI coding wars with OpenAI.