Anthropic Releases Claude 4 with 2-Million Token Context Window
Anthropic's latest flagship model extends context length to 2 million tokens — equivalent to roughly 20 full-length novels — enabling entirely new use cases in legal, scientific, and enterprise AI.
Visualization of large-scale AI language model context processing
Anthropic has released Claude 4, the latest version of its flagship AI assistant, featuring a 2-million token context window that the company says is the largest available in any commercially released model.
What 2 Million Tokens Means in Practice
At 2 million tokens, Claude 4 can process approximately 1,500 pages of text, 20 full-length novels, or an entire software codebase — in a single prompt. This unlocks use cases that were previously impossible: analysing an entire legal case file, reviewing a complete medical record history, or refactoring a large application at once.
Performance Benchmarks
Anthropic claims Claude 4 significantly outperforms its predecessor on the Needle in a Haystack benchmark — a test of long-context recall — achieving near-perfect scores even at the maximum context length. This has historically been a weakness of large context models.
Pricing
Claude 4 is priced at $15 per million input tokens and $75 per million output tokens — approximately 3x the cost of Claude 3.5 Sonnet, reflecting the computational demands of the larger context window.
Safety Measures
Anthropic has introduced new safety evaluations specifically for long-context models, addressing concerns that models with very large context windows could be more susceptible to prompt injection attacks embedded in long documents.
