Claude upgrades to 1 million token context window for free
In a move considered a 'game-changer' in the AI industry, Anthropic has just announced a revolutionary upgrade: offering a 1 million token context window for Claude Opus 4.6 and Sonnet 4.6 at standard pricing. Notably, there are absolutely no surcharges for long context, a policy completely opposite to most other AI providers, including Google and OpenAI, who typically increase prices for larger context limits.How much does Claude with a one million token context window cost?This is a massive shift brought by Anthropic. One million tokens is equivalent to about 750,000 words, enough to process 10 to 15 full novels in a single run. This number has immense significance in real-world workflows:Instead of having to chunk codebases or documents into parts, users can input an entire project into a single session, allowing the AI to work with it as a unified whole.Claude can retain all information from start to finish when analyzing thousands of pages of legal contracts, eliminating the risk of losing context halfway.Complex long-context processing techniques previously required, such as document chunking, lossy summarization, or clearing context to free up memory, are now no longer necessary.New pricing structure with no surcharge for one million tokensThe most surprising aspect is the new pricing structure, which completely lacks any long-context surcharge. The standard pricing applies to the entire range from 1 to 1 million tokens:Opus 4.6: $5 input and $25 output per 1 million tokens.Sonnet 4.6: $3 input and $15 output per 1 million tokens.For perspective, previously, when using a context window exceeding 200,000 tokens, many providers typically charged 2 to 4 times more. Particularly, Claude Code users with Pro ($20), Max ($100), Team, and Enterprise plans automatically receive the 1 million token context window when using Opus 4.6 without needing extra usage credits.Does Claude Opus 4.6 really remember all 1 million tokens?A common question when increasing the context window size is whether reasoning quality is compromised. Anthropic addressed this concern with impressive benchmark results.Claude Opus 4.6 scores 78.3% on MRCR v2 – a test measuring the ability to retrieve hidden information within a massive amount of text. This is the highest score among all current advanced models at the same context length. For comparison:GPT's accuracy drops significantly, reaching only 36% at 1 million context.Gemini performs even worse, at just 26%, showing that these models 'forget' up to ⅓ of what was provided previously when reaching long context.With Sonnet 4.6, the ability to remember over ⅔ of a long context further proves Anthropic's leading position in processing complex information. Claude's media limit increases 6x to 600 images per requestAlongside the context window, another upgrade that received less attention but is extremely important is the media limit. This limit has increased to 600 images or PDF pages per request, 6 times higher than the previous 100. This is especially meaningful for those working intensively with tasks requiring the processing of many visual documents or PDFs, which is truly significant for Claude Pro and Max users who constantly hit token limits and have to wait.Significantly reducing usage costs for businessesThis breakthrough feature is available immediately on the Claude Platform, Microsoft Azure Foundry, and Google Cloud Vertex AI. For Claude Code users on Max, Team, and Enterprise plans using Opus 4.6, the 1 million token context window is enabled by default without additional setup. This not only improves performance but also significantly reduces costs for AI systems that frequently call Claude's API, bringing major economic benefits to businesses and developers.