GPT-5.6 Sol, 3-tier preview puts OpenAI model access under federal review

A frontier launch with a policy gate

OpenAI’s latest model news is not just a benchmark update. On June 26, 2026 at 17:10:20 UTC, the company posted that GPT-5.6 is entering a limited preview with three tiers: Sol as the flagship model, Terra for balanced everyday work, and Luna for high-volume tasks. FxTwitter recorded more than 16.2 million views, 38,000 likes, and 5,500 reposts during collection, making it the strongest public AI product signal inside the 48-hour window.

“Introducing a limited preview of GPT-5.6 Sol … Terra … and GPT-5.6 Luna.”

OpenAI’s account usually carries first-party product launches, safety notes, and developer updates, so this tweet is the primary pointer to the rollout. The linked OpenAI page adds the more consequential detail: access starts with a small group of trusted partners whose participation was shared with the US government. OpenAI says broader availability is planned in the coming weeks, but also says this kind of government access process should not become the long-term default.

The technical claims are substantial. OpenAI says Sol is its strongest model yet, adds a max reasoning effort, and introduces an ultra mode that uses subagents for complex work. The page cites state-of-the-art performance on Terminal-Bench 2.1, stronger biology workflow results on GeneBench v1, and cyber performance where Sol is competitive with Mythos Preview while using about one third of the output tokens. Pricing also changes the product map: Sol is $5 input and $30 output per 1M tokens, Terra is $2.50 and $15, and Luna is $1 and $6.

What to watch next is whether the limited preview becomes a release template for powerful models. If government review, partner eligibility, and cyber safeguards become part of ordinary model availability, developers will have to track access policy as closely as latency, price, and benchmark scores. Source: OpenAI source tweet · OpenAI model preview