Anthropic's introductory $1 offer to the US government was due to expire on 31 October 2026, and its replacement shows where government AI pricing is heading. Under an expanded OneGov agreement with the General Services Administration, announced on 27 October, government buyers get 40% off Claude Enterprise, Claude for Government and the Claude API. For the first time the discount covers every way of buying Claude, and it reaches state, local, tribal and territorial governments as well as federal agencies.
The terms are simple. There is no minimum spend; agencies pay for what they use at the reduced rate. Buying direct from Anthropic is available now through GSA's Multiple Award Schedule, and purchases through cloud providers are promised in the coming months. Anthropic puts the expected saving to government at about $1 billion. It says about 500 US government organisations already use Claude, weekly active users of Claude for Government have grown about 8 times since February, and use across federal civilian agencies has grown about 20 times in a year.
The case studies lean on Claude Code. In Utah, 34 developers across 10 state agencies tried it in a four-week pilot; 80% had it working within 30 minutes and 71% said they saved more than 20 hours. An unnamed state IT developer moved almost 900 API methods to a new framework in two days against a planned month, and an unnamed federal security team used Claude Mythos to find and patch a serious flaw that a human penetration test had missed a month earlier. Claude for Government holds FedRAMP High authorisation, as do Claude models on Amazon Bedrock and Google Cloud Vertex AI, while Claude Enterprise is working towards FedRAMP Moderate.
The move fits a pattern. In August 2025, GSA struck nominal-price OneGov deals with Anthropic, OpenAI and Google, the last at 47 cents per agency. Those offers are now giving way to discounts on metered use, and ExecutiveGov reports that OpenAI's own 27-month OneGov deal gives agencies 50% off token-based ChatGPT usage. A larger percentage does not settle which is cheaper, because list prices and the tokens each model uses for a task differ. For readers comparing the Claude plans we review, the discount narrows the gap with rivals on the API, where frontier models such as Claude Mythos 5.2 are expensive at list price.
Agencies should treat the announcement with some care. The usage figures and every case study come from Anthropic, most of them unnamed, and none has been independently checked. The post gives no end date for the agreement, so the price after it lapses is unknown. Deep discounts with no commitment make it easy to start and harder to leave once workflows are built around one vendor, which is why a total-cost comparison across the OneGov offers matters more than the headline percentage.