Anthropic Expands Claude Context Window to 100K Tokens

Anthropic has dramatically increased the context window of its Claude large language model from 9,000 tokens to 100,000 tokens. This expansion allows the system to process approximately 75,000 words of text in a single prompt. While a human reader typically requires more than five hours to read and digest this volume of information, Claude can analyze the entire dataset in less than 60 seconds.

To demonstrate this capability, Anthropic loaded the complete text of The Great Gatsby, totaling 72,000 tokens, into its Claude-Instant model. The team altered a single sentence to describe the character Mr. Carraway as "a software engineer that works on machine learning tooling at Anthropic." When prompted to identify the modification, Claude-Instant successfully located the change in just 22 seconds.

For developers and enterprise users, this upgrade changes how businesses interact with large documents. Instead of relying on traditional vector search methods, which can struggle with complex, multi-part synthesis, practitioners can now feed entire financial statements, legal files, or developer documentations directly into the API. Anthropic notes that the 100,000-token limit also accommodates roughly six hours of audio. In one test, partner AssemblyAI transcribed a long podcast into 58,000 words and used Claude to perform summarization and question-answering tasks.

The new 100K context window is currently available to partners through the Anthropic API. Alongside this release, the company announced a 5 million dollar grant program to fund independent research into AI's impact on human wellbeing, and introduced updates to Claude Fable 5 to reduce false positives in its biology safeguards.