OneGov ChatGPT Pricing Shifts to Metered Usage as OpenAI Ends Its $1 Federal Deal
The era of near-free federal ChatGPT is over. OpenAI and the U.S. General Services Administration have replaced the $1-per-agency pilot with a metered OneGov ChatGPT pricing model that gives federal, state, local and tribal agencies 50% off standard token rates, no platform-access fee and no minimum spend commitment.
The agreement was announced on September 10, 2026 and takes effect October 1, running through December 31, 2028, a 27-month term. It supersedes the OneGov partnership unveiled in August 2025, under which participating agencies received ChatGPT Enterprise for a nominal $1 for a year. Promotional pricing under the GSA's Multiple Award Schedule, which also covers Gemini for Government and Anthropic's Claude, lapses on September 30, and access does not continue automatically.
What the New OneGov ChatGPT Pricing Changes
Cost was previously decoupled from consumption. An agency paid a dollar and drew on the model with no usage meter attached. The replacement structure ties the invoice to tokens consumed, billed in U.S. dollars, at half the commercial rate across ChatGPT models, including deployments inside FedRAMP-authorized environments.
FedRAMP authorization matters for the expansion that follows. It lets agencies run the models inside environments they have already accredited, which removes a security review that would otherwise sit between a signed agreement and actual use. Daybreak Blue applies the same logic to cyber defence, where OpenAI is positioning itself alongside established government security vendors rather than purely as a productivity supplier.
Coverage widens at the same time. OpenAI and the GSA are extending the arrangement to roughly 23 million U.S. public-sector employees across federal, state, local and tribal governments. Verified government entities also qualify for Daybreak Blue, OpenAI's advanced cyber-defender product, at 50% off standard commercial pricing with training and enablement included.
The jump from 3.5 million federal users to a public-sector pool of roughly 23 million is the largest structural change in the arrangement. It moves the contract from a Washington pilot into a procurement vehicle that reaches county administrations, municipal agencies and tribal organisations, each with its own budget cycle and its own tolerance for a variable bill.
The fine print shapes any spend forecast. OpenAI's enterprise rate card states that the promotion covers token-based usage and excludes legacy metering. Included plan usage, five-hour and weekly limits, and legacy credit rates are unchanged. The discounted rates apply to ChatGPT Work and Codex activity, a set that spans local tasks, cloud tasks, automations, code review, auto review and delegated workers.
| Dimension | Original OneGov pilot (August 2025) | New agreement (October 2026 onward) |
|---|---|---|
| Licence fee | $1 per agency, per year | $0 platform-access fee |
| Billing basis | Flat pilot licence | Metered, per-token usage in USD |
| Rate | Nominal $1 | 50% off standard commercial pricing |
| Minimum spend | Not specified | None |
| Term | One year | October 1, 2026 to December 31, 2028 |
| Eligibility | Participating federal agencies | Federal, state, local and tribal governments |
The Arithmetic Behind a Half-Price Rate
A 50% discount on tokens is not half of what agencies paid before. Before, they paid about a dollar per agency. The relevant comparison is between the new metered bill and zero, which turns the discount into a commercial concession rather than a budget line that shrinks.
The GSA's own accounting supplies the counterweight. The agency reports that ChatGPT was accessible to 3.5 million federal employees during the pilot and attributes $1.4 billion in cost savings to ChatGPT and other AI models procured through OneGov. Those savings accrued under flat pricing, where the marginal cost of a prompt was invisible to the individual agency that issued it.
Metered billing restores that signal, and with it an incentive to consolidate overlapping use cases, retire automations that never paid for themselves and route routine work to cheaper models. It also transfers forecasting risk to the buyer. A fixed annual licence is simple to plan and simple to waste; consumption billing is difficult to plan and difficult to waste.
Agencies that cannot estimate token volume before October 1 will meet their number on the first invoice instead of in a planning cycle. The spending exposure is concentrated in exactly the teams that leaned hardest on the pilot: policy drafting, contract review, data analysis and software maintenance.
Procurement structure shapes behaviour as much as price does. A vehicle with no minimum spend and no platform fee lowers the cost of starting, and it also removes the volume concessions that usually come with committed spend. An agency buying a few dozen seats pays the same 50% rate as a cabinet department buying for thousands of staff.
The Vendor Field and the Lock-In Question
OneGov is not an OpenAI-only programme. The GSA has been assembling model vendors under the same banner, and the schedule terms covering Gemini for Government and Claude expire at the end of September alongside the original ChatGPT arrangement. OpenAI now holds a defined runway through the end of 2028 while rival promotional terms lapse without an announced successor of comparable length.
That asymmetry is the strategic core of the deal. The longer an agency's workflows sit inside ChatGPT, the higher the cost of moving them elsewhere, and the harder it becomes to treat the model as a commodity to be re-bid. The GSA attributes $1.4 billion in savings to models procured through OneGov broadly, but the multi-year paper commitment belongs to a single vendor.
OpenAI's wider pricing behaviour points the same way. ChatGPT Business costs $20 per user per month on annual billing and $25 monthly, a $5 reduction since April 2, 2026, which prices a governed team seat at the level of an individual Plus plan. Nonprofits can obtain up to 75% off ChatGPT Business or Enterprise. OpenAI's head of ChatGPT has signalled that unlimited plans may not survive as the company's cost structure changes.
The direction is consistent: away from flat, headline-friendly pricing and toward consumption-linked billing, with discounts used to hold strategic accounts while usage becomes the revenue engine. A separate OpenAI move in August 2026 cut API pricing for its GPT-5.6 Sol model by 20% for three months, and that promotion touched metered and API billing only, leaving Pro, Plus and Business subscription usage untouched.
What Agencies Should Do Before October 1
Three tasks follow from the terms. The first is measurement, because the pilot never produced per-team usage data, so historical consumption has to be reconstructed and converted into token estimates at discounted rates. The second is governance, since the discount applies to rates rather than to volume, which means the real lever is deciding which workloads deserve frontier-model tokens and which can run on cheaper tiers.
The third is contract hygiene. Access stops rather than continues when the September 30 schedule terms expire, so any agency that has not signed under the new vehicle loses its seats regardless of how much it would have been willing to pay.
Vendors face a parallel calculation. A 50% federal discount becomes an anchor price that commercial buyers will cite in negotiations, and with nonprofit pricing already reaching 75% off, the ladder of discounts is legible to procurement teams well beyond Washington.
The open question is behavioural rather than financial. Federal employees who built daily habits around ChatGPT during a year of near-free access are unlikely to drop them, so the metered bill will be driven by entrenched usage rather than fresh adoption. Whether that yields a manageable invoice or an awkward budget hearing depends on how quickly agencies build measurement discipline the pilot never required.
Why this matters
The switch from a $1 licence to metered billing turns a symbolic pilot into a recurring procurement line, and it makes federal AI spending visible in a way the flat arrangement never did. For agencies, the discipline that follows is measurement: knowing which workflows consume tokens and which do not. For OpenAI, it is a bet that entrenched daily use will outlast a discount designed to create it.
Sources
ChatGPT Rate Card (Enterprise token-based pricing) | OpenAI Help Center
Photo by Brecht Corbeel on Unsplash
Related Articles
- GPT-5.6 Sol price cut: OpenAI's 3-month defense
- GPT-6 Astra Pricing: OpenAI Opens Flagship Model to Enterprises
- GitHub Copilot Moves to Usage-Based Billing for Enterprises
✔Human Verified
Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.