Claude gets 40% off for every level of US government

Anthropic on October 27 signed a new OneGov agreement with the General Services Administration that takes 40% off Claude Enterprise, Claude for Government and the Claude API for US government buyers, with no minimum spend and billing only for the usage consumed. Agencies buying directly from Anthropic get the discount now, while those running Claude through cloud providers get it only “in the coming months,” Anthropic said in its announcement.

The deal reaches past federal agencies to state, local, tribal and territorial governments, and purchases go through GSA’s Multiple Award Schedule. California’s earlier deal, announced by Gavin Newsom, gave its agencies Claude at half price, a deeper cut than this one. Anthropic estimates the government’s savings at $1 billion and says about 500 federal, state and local government organizations use Claude.

Direct now, cloud later

Which route an agency uses decides when the 40% applies. Claude for Government became generally available on September 30 at FedRAMP High, the federal cloud authorization level for sensitive unclassified work, and Anthropic says it offers near-feature parity with its commercial products, including Claude’s coding and agentic work capabilities. Agencies building their own tools reach Claude through Amazon Bedrock, where it is authorized for FedRAMP High and select models for DoD IL4 and IL5 workloads in AWS GovCloud, or through Google Cloud’s Vertex AI, also at FedRAMP High. Those cloud routes are the ones still waiting for the discount.

Claude Enterprise is in process for FedRAMP Moderate through the FedRAMP 20x program. Anthropic is also launching a dedicated program to help agencies implement Claude, built on what its private-sector customers already use, in the coming weeks.

Paying by consumption

FedScoop places the deal right after OpenAI’s switch to consumption-based pricing the month before, a model it says has rapidly replaced seat-based pricing in private-sector enterprises and is spreading to the federal government. The outlet also carries the budgeting worry: Gartner said such pricing makes costs difficult to predict because vendors lack transparency into how token consumption is calculated and billed, and Forrester’s Tracy Woo wrote in August that “a small group of power users can drive a disproportionate share of costs, turning successful AI adoption into both a productivity win and a budgeting challenge.”

OpenAI’s terms, announced on September 10, give 50% off token usage, a $0 standard ChatGPT seat fee, no minimum orders and no spend commitment, run from October 1, 2026 through December 31, 2028, and are open to legislative and judicial buyers plus state, local and tribal governments. Anthropic’s 40% is the smaller cut, though OpenAI’s figure is stated for token usage and Anthropic’s spans Claude Enterprise and Claude for Government as well as the API.

On Claude for Government, the unpredictability Gartner describes had a ceiling at launch: the platform billed usage in fixed increments under a hard not-to-exceed cap, so spending could not go past what an agency had obligated, with no seat fees.

The new pricing lands four days before Anthropic’s $1-per-user OneGov offer runs out. A listing on GSA’s OneGov website extended that offer through October 31 instead of September 30, Nextgov reported.

Anthropic’s own numbers

Weekly active users of Claude for Government are up about eightfold since February, and usage across federal civilian agencies has grown about 20-fold in a year, Anthropic says. Its most detailed case is Utah, which gave Claude Code to 34 developers across 10 state agencies: 80% had a working environment within 30 minutes, and 71% reported saving 20 or more hours by the end of the four-week pilot.

The other examples are unnamed. “A US state CIO” credits Claude with responding to a ransomware attack in 8 minutes against the usual 2-3 days. A state IT developer moved nearly 900 API methods to a new framework in two days on a project scoped for a month. And a federal security team used Claude Mythos to find and patch, in about two days, a vulnerability that would have allowed the takeover of a critical production system. A human penetration test the month before had missed it.