OpenAI is ending the $1 per year pilot that gave US federal agencies its models and moving them to usage-based pricing at a 50% discount, the General Services Administration said on 10th September, 2026.
Key Points
- OpenAI will replace its $1 per year federal pilot with usage-based pricing set 50% below standard prices.
- The General Services Administration said 3.5 million federal employees used ChatGPT during the pilot.
- The GSA says discounts on AI tools across its OneGov program have saved the government $1.4 billion.
- The new agreement takes effect on October 1 and runs for more than two years.
- OpenAI plans to extend the discount to state, local, and tribal governments.
OpenAI moves agencies onto metered pricing
OpenAI joined OneGov, the GSA marketplace that lets federal agencies buy AI software at scale, in August 2025. The original GSA announcement set the first-year price at $1. According to Bloomberg, those terms were the steepest tech discount the government’s central purchasing arm had ever negotiated.
The replacement deal raises the price, but the GSA framed the 50% rate as a way for agencies to test and scale AI tools. That includes OpenAI’s new GPT-6 Astra model. OpenAI is set to announce the expansion beyond federal buyers on Thursday.
OpenAI eliminates its symbolic one dollar per year agreement for federal entities, replacing it with a structured fifty percent discount on its premium artificial intelligence models for US government agencies.https://t.co/g2V1uF3Ztz
— NDTV Profit News (@NDTVProfit) September 10, 2026
Anthropic’s matching deal ran into the Pentagon
Anthropic PBC matched the $1 per year price days after OpenAI’s 2025 deal. A few weeks later, the Claude maker hosted an event for federal workers at Washington’s Union Station, featuring an in-person speech by CEO Dario Amodei.
That push stalled when Anthropic sought guardrails on surveillance and autonomous strikes in national security contracts. Defense Secretary Pete Hegseth labeled the company a supply-chain risk. President Donald Trump then used Truth Social to order all agencies to stop using Anthropic.
Anthropic sued. The GSA first moved to pull Anthropic from its central AI model platform, then reversed course after a preliminary injunction in the California case. Last month, the judge in that case ruled the administration must lift its ban for federal agencies, suggesting the Pentagon can bar the technology from its own work but may lack authority to force other agencies to drop it.
OpenAI said in February it had reached an agreement with the Pentagon to provide its tools.
Several questions remain unresolved:
- Will the Washington case, where a three-judge panel has yet to rule on a separate legal argument, reach a different outcome?
- Will civilian agencies without a national security focus keep using Claude once both lawsuits end?
- Which standard prices anchor the 50% discount, and what will a typical agency pay?
Why It Matters?
Agencies built daily habits around ChatGPT at almost no cost. Metered pricing now ties spending to how heavily staff use the tools. Half off standard prices is still a steep public sector discount, and the move to state, local, and tribal buyers widens OpenAI’s footprint while its closest federal rival’s status sits in court.
The next marker is the October 1 start date. Agency IT and procurement teams already running ChatGPT should review current usage volumes before then, since costs will scale with use under the new terms. State and local buyers should check OpenAI’s announcement for their own terms.
