Uncle Sam Approved: OpenAI Drops GPT-5.6 Sol This Thursday After Secret Vetting
Get ready to chat with a neural network that has been officially certified safe by Washington suits. Because nothing says cutting-edge technology quite like a bunch of politicians inspecting lines of code for two weeks.
The US Department of Commerce finally gave OpenAI the green light to unleash its new GPT-5.6 family, ending a two-week period where only twenty hand-picked corporate giants could play with the model under strict government supervision.
To get this clearance, OpenAI engineers had to camp out in Washington, playing twenty questions with the AI Safety Standards and Innovation Center. Apparently, government officials wanted to make sure the AI wouldn't accidentally teach citizens how to tax-evade, leading to a hilarious client-by-client vetting process for the early preview.
The star of the show is GPT-5.6 Sol, a heavy-duty flagship model designed for complex multi-step reasoning, boasting fancy sub-agent modes. It will cost $5 per million input tokens and $30 per million output tokens, which is basically peanuts for corporations but might hurt the wallets of indie developers trying to build yet another wrapper app.
For the budget-conscious, OpenAI is also dropping Terra, which matches the power of the older GPT-5.5 but at half the price, and Luna, a lightning-fast model designed to be dirt cheap. To make things even faster, OpenAI is hosting Sol on massive Cerebras chips, pushing speeds up to a mind-melting 750 tokens per second for a few chosen VIPs.
This whole bureaucratic dance isn't even a formal law yet, as both OpenAI and its rival Anthropic—which had its own Fable 5 model vetted last month—are just playing along with the White House's upcoming AI rules under Trump. The models are expected to go live around 8:00 PM.
We have officially entered the era where math equations require federal clearance before they can write poetry or debug Javascript. Soon, developers will need a background check just to run a local LLM on gaming rigs, while regulators pretend they actually understand how deep learning works.
Source: Axios
Comments
This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.