The company is working closely with the Pentagon’s Chief Digital Artificial Intelligence Office to deliver more model access via GenAI.mil. #
Speaking during the Defense One Tech Summit in Virginia on Tuesday, Mohammed Husain — the Strategic Delivery Lead for Cyber at OpenAI — said the company is poised in early July to unveil its flagship chatbot model ChatGPT to defense civilian and military personnel through GenAI.mil, the Pentagon’s enterprise-wide generative AI platform.
Through GenAI.mil, ChatGPT will be available to more than 3 million defense personnel and certified for controlled unclassified information and Impact Level 5. Husain said OpenAI is still working closely with the Department of Defense’s Chief Digital Artificial Intelligence Office to launch ChatGPT across the military apparatus.
The Defense Department launched the platform in December with initial plans to integrate Gemini for Government, and later announced plans to incorporate AI models from OpenAI and xAI. In late April, senior defense officials said more than 1.3 million users were regularly using the platform, having developed more than 100,000 AI agents.
Husain forecasted the customized version of the chatbot will make its debut in “the coming weeks.”
“I think we're going live extremely soon, and excited to make a broader announcement about that in early July,” he said.
Federal agencies have been using ChatGPT since at least January 2025, and the company offered its model at a significant discount through a OneGov deal with the General Services Administration in August. OpenAI’s latest model, ChatGPT 5.4, was made available to the federal workforce on Amazon’s Bedrock and overarching GovCloud platforms as of earlier this month.
Husain forecasted demand for both higher volumes of and more efficient tokens, which are converted data able to be interpreted and processed by an AI system.
“These models consume a ton of tokens, and it turns out that if you want to complete the most valuable work, it's going to take more tokens,” Husain said. “And so one thing I think will become much more a part of the conversation … is this concept of token efficiency.”
Husain likened token efficiency to being less about compute processing speed, but more about cost effectiveness per completed task. He referenced the news from early June, in which OpenAI’s GPT-5.5, GPT-5.4 and Codex were made available on Amazon Bedrock, saying that this enables further deployment of more intelligent, token-heavy models.
“I think deploying these models, they're going to be much more intelligent, they're going to consume more tokens,” Husain said. “So I think cost efficiency is going to become a really interesting part of the story.”
Compute infrastructure is still a consistent topic for partners in the government space, Husain said, for both multicloud and on-prem environments, a void companies like AWS are hurrying to fill.