Anthropic has new pricing terms with Washington. On October 27 the company and the General Services Administration, the agency that negotiates purchasing terms on behalf of federal buyers, said they had widened their OneGov agreement for Claude. Government customers now get 40% off usage-based prices for the Claude API, Claude Enterprise and Claude for Government, and the offer reaches past federal agencies to state, local, tribal and territorial governments.
The structure matters as much as the size of the cut. According to Anthropic, this is the first time its government discount covers every product surface: agencies buying directly qualify immediately, and purchases made through cloud providers are due to follow in the coming months. GSA describes the deal as consumption-based with no minimum order, so an agency is billed for the tokens it actually uses. Orders can go through GSA's Multiple Award Schedule, through resellers or through major cloud platforms, the agency said.
The agreement replaces a promotional arrangement that is about to run out. Under the earlier OneGov offer, federal agencies could use Claude's enterprise and government products for $1 per user, and that deal ends on October 31, Nextgov/FCW and FedScoop reported. The shift follows OpenAI, whose own token-based OneGov agreement, carrying a 50% discount over 27 months, took effect at the start of October, according to Nextgov.
Anthropic puts the expected savings at about $1 billion, a figure it has not broken down. It also published adoption numbers that have not been checked independently: roughly 500 US federal, state and local government bodies use Claude, weekly active users of Claude for Government have grown about eightfold since February, and use across federal civilian agencies is up around 20 times in a year. Among customer examples, the company cites a Utah pilot in which 34 developers from 10 state agencies tried Claude Code for four weeks, and a federal security team that it says used Claude Mythos to find and fix a serious flaw a human penetration test had missed a month earlier. Anthropic also said it will launch a dedicated implementation program for agencies in the coming weeks.
Security clearances are the other half of the pitch. Claude for Government, generally available since September 30, runs at the FedRAMP High level, Anthropic said. Claude carries a FedRAMP High authorization on Amazon Bedrock, with some models cleared for Defense Department IL4 and IL5 workloads in AWS GovCloud, and on Google Cloud's Vertex AI it holds FedRAMP High. Claude Enterprise is still pursuing FedRAMP Moderate under the FedRAMP 20x program.
The timing is notable because Anthropic remains in a legal dispute with the Pentagon, as FedScoop pointed out, while its ties with civilian agencies look steady. GSA said OneGov has produced more than $1.8 billion in savings since it started, about $1.6 billion of it from AI agreements. Google's current Gemini offer for agencies, priced at $0.47 per agency, runs until November 15, and no follow-up terms have been disclosed, FedScoop reported.
Metered pricing brings its own risk. Analysts at Gartner and Forrester, quoted by FedScoop, have warned that token billing makes spending hard to forecast because a small group of heavy users can drive much of the cost. Agencies moving from a flat $1 seat to pay-per-use may find their bills rise with adoption, and the $1 billion savings estimate is Anthropic's projection rather than an audited result.