# Claude's GSA OneGov Discount: 40% Off, Eligibility and Cloud Rollout

The new government agreement expands usage-based pricing, while ordering channels and security requirements still need separate checks.

By Paras Madan | 2026-10-28


Anthropic and the US General Services Administration announced an expanded OneGov agreement on October 27 with a 40% discount on eligible Claude usage and no minimum purchase. The offer broadens access beyond federal agencies, but buyers should distinguish today's direct purchasing route from the cloud-provider rollout still ahead.

## What the discount covers

[GSA's announcement](https://www.gsa.gov/about-gsa/newsroom/news-releases/gsa-anthropic-reach-onegov-agreement-on-discounted-consumptionbased-pricing-10272026) describes discounted token-based consumption across Claude models available to government customers. Agencies pay for what they use without a minimum order. Federal executive, legislative and judicial bodies, as well as state, local and tribal governments, are included in its eligibility description.

[Anthropic's announcement](https://www.anthropic.com/news/expanding-partnership-gsa) names Claude Enterprise, Claude for Government and the Claude API. It also explicitly includes territorial governments. This is a government purchasing arrangement, not a 40% cut to everyone's Claude subscription or a new public API list price.

For a simple illustration, $1,000 of eligible usage at the applicable undiscounted rate would become $600 at a 40% discount. That calculation says nothing about the size of an agency's eventual bill, which still depends on consumption and its ordering terms.

## Direct access now, cloud details still matter

There is a timing distinction between the two announcements. GSA describes direct, reseller and major-cloud purchasing routes broadly. Anthropic says direct purchases receive the discount today, with cloud-provider purchases following in the coming months.

**Do not assume an existing Bedrock or Vertex AI bill automatically receives the new rate.** The practical next step for an agency is to verify its chosen channel, eligible models and effective date with the ordering team.

The public [OneGov directory](https://itvmo.gsa.gov/onegov/) also still showed the earlier $1-per-user Anthropic offer, expiring October 31, in the version reviewed for this article. The new dated announcement is clearer about the expanded consumption arrangement, but a directory entry should not be treated as a complete current quote. Detailed ordering materials are available through the government-only portal.

## What changes from the earlier government product

Claude for Government itself is not launching this week. It became generally available on September 30. Its [product documentation](https://claude.com/resources/articles/claude-for-government-is-now-generally-available) describes usage billing without seat fees, a hard spending cap and department-level administration.

Administrators can set group limits, allocate budgets and inspect usage by model and user. The platform also provides audit logs and supporting documentation for an agency's authorization process. Those controls matter when a small pilot becomes a service used across departments.

The October agreement changes the economics and eligibility around that use. An agency can test a narrow process without committing to a minimum purchase, then make a better-informed decision about expanding it.

## A discount does not settle the deployment decision

Anthropic describes Claude for Government as FedRAMP High authorized. It separately says Claude Enterprise is in the process of seeking FedRAMP Moderate authorization through FedRAMP 20x. Those statements refer to different offerings and should not be collapsed into one blanket certification claim.

A buyer still needs to match the environment to the data and workload. Pricing eligibility alone does not answer which integrations are allowed, who can approve an action or where a department's records can be processed.

The same distinction applies to savings. A lower token rate reduces one component of a workflow's cost. Integration, review and training can remain substantial. If an agent creates more work for staff to check, a cheaper model call will not necessarily make the process cheaper overall.

## What agencies should measure next

A useful pilot would track cost and staff time per completed case, with the same quality threshold before and after deployment. Spending limits make that experiment easier to contain, while the absence of a minimum order makes it easier to stop if the result is weak.

The more important follow-up will be whether agencies can move from discounted access to repeatable work that survives normal operational review.

I'm Paras. I work at the intersection of AI, engineering and distribution.

I help teams explain complex enterprise AI products clearly to technical buyers and decision-makers.

[Book a 30-minute call](https://cal.com/paras-madan/30min).