Anthropic has announced two major updates for its Claude AI models — a huge jump in how much information they can handle at once, and a special offer for the US government that matches and expands on a recent move by OpenAI. The company said it will make its Claude models available to all three branches of the US government — executive, legislative, and judiciary — for just $1 per year. This offer will be valid for one year and is aimed at making Claude more widely used in public sector work. It comes just a week after OpenAI revealed that ChatGPT Enterprise would be offered to the entire federal executive branch for the same $1 yearly fee. Anthropic’s deal goes further by targeting every branch of government, not just the executive arm.

Longer memory for Claude Sonnet 4

Alongside the government push, Anthropic is rolling out a major improvement to its Claude Sonnet 4 AI model. The model’s context window (the amount of information it can process in one go) has now been increased to 1 million tokens. That’s roughly the same as 750,000 words or about 75,000 lines of code. To put it another way, the model can take in more than the entire Lord of the Rings trilogy in a single prompt.

This new limit is five times larger than Claude’s earlier 200,000-token capacity and also more than double the 400,000-token limit of OpenAI’s GPT-5. The upgrade will make Claude more capable for developers and enterprises that need to work with large amounts of data at once, such as analysing massive codebases or processing long research documents.

Enterprise customers will be able to use this long context window through Anthropic’s API. The feature will also be available through its cloud partners, including Amazon Bedrock and Google Cloud’s Vertex AI, making it accessible to organisations already working on those platforms.

Focus on government with strong security standards

Anthropic will provide two main versions of its AI for customers — Claude for Enterprise and Claude for Government. The government-specific offering is built to support “FedRAMP High” workloads, which is the strictest security standard for handling sensitive but unclassified US government data.

This means federal employees can use Claude for tasks that involve confidential information while staying within official security requirements. Anthropic says Claude can help with a variety of government-related work, from accelerating scientific research to improving citizen services. For example, the model can be used to summarise complex documents, translate information into multiple languages, or assist in technical projects.

The company also plans to give agencies technical support so they can integrate Claude into their existing systems. Because of partnerships with AWS, Google Cloud, and Palantir, agencies can use Claude through their current secure infrastructure without making major changes to their setup.

Part of a bigger AI competition

This move is the latest in an ongoing battle between major AI companies to gain a bigger role in government projects. Recently, the US General Services Administration added Anthropic, OpenAI, and Google DeepMind to its list of approved AI vendors for civilian agencies. Being on this list makes it easier for government bodies to buy their services.

OpenAI has already made its own $1-a-year offer to the executive branch, and industry watchers are now waiting to see whether Google will respond with a similar deal.

Anthropic is already involved in government-linked AI work. It is one of several companies, along with OpenAI, xAI, and Google, that have been granted up to $200 million by the US Department of Defense to use AI for national security purposes. But the company’s latest announcement suggests that it is looking beyond defence and into areas like scientific research and public health.

– Ends