GPT-5.6 Sol, Terra, and Luna: confirmed pricing and what the staged rollout means for your Q3 stack
Sol at $5/$30, Terra at $2.50/$15, Luna at $1/$6 per million tokens. OpenAI launched GPT-5.6 in limited preview on June 26 with access staged by White House request. The pricing is real and plannable now. The access is not yet open.
The pricing for GPT-5.6 is confirmed. Sol, the flagship tier, runs $5 input / $30 output per million tokens — roughly in line with Opus 4.8 for the hardest tasks. Terra, the mid-tier, is $2.50 input / $15 output; OpenAI describes it as GPT-5.5 Pro performance at approximately half the cost. Luna, the fast volume tier, is $1 input / $6 output, in the range of Sonnet 4.6 for simpler pipeline steps.
What you cannot do is access any of them yet. OpenAI launched the preview on June 26, restricted to approximately 20 government-approved organizations via API and Codex. ChatGPT integration follows at general availability. OpenAI is targeting early July.
That gap between "pricing confirmed" and "access open" is worth understanding before GA arrives.
Why the rollout is staged
On June 25, the Office of the National Cyber Director and the Office of Science and Technology Policy asked OpenAI to hold the broader launch. GPT-5.6's capabilities — the models are particularly capable at vulnerability identification and exploitation — triggered a request for staged access. Sam Altman confirmed to employees that the administration would be approving organizations one by one before a wider rollout.
The framework behind this is Executive Order 14409, signed June 2, which establishes a voluntary arrangement where AI developers can offer the federal government up to 30 days of pre-release access to evaluate frontier models before public deployment. The August 1 deadline in the EO is for agencies to have the evaluation framework operational, not the end of any specific review window.
The Fable 5 suspension, in place since June 12, operates through export control authority rather than EO 14409. But the builder experience is the same: a model you planned around is not accessible on the timeline you expected. Two of the most capable AI models in existence are gated for general developers in the same month under related but distinct government authority. The legal mechanisms differ; the operational consequence does not.
Planning your stack before GA opens
Update your cost models now. Terra at $2.50 / $15 is cheaper than the current GPT-5.5 Pro tier at both input and output. If Terra delivers on OpenAI's description when general access opens, that is a clear migration path for production pipelines currently on GPT-5.5 Pro. The pricing is confirmed enough to model against; the performance claim requires testing. Your production baseline stays on GPT-5.5 Pro or Sonnet 4.6 until you can run that test — but there is no reason to wait until GA to prepare the comparison.
If you have Fable 5 workflows on hold: no change this week for general access. Mythos 5 was partially restored on June 27 for a narrow set of entities — US critical infrastructure operators on the government's Annex A list, their foreign contractors, and US government agencies and national laboratories. Fable 5 remains suspended globally with no announced general availability date. The next structural date in the ongoing negotiations is July 8. Opus 4.8 and Sonnet 4.6 remain the operational stack.
If you built custom GPTs on GPT-4.5: that model retired from ChatGPT on June 26. Existing conversations migrated to GPT-5. Verify your custom GPTs are behaving as expected under the new model — output style and latency both differ. GPT-4.5 remains accessible via API.
The thing worth updating is your dependency model, not just your release calendar. A production stack that is tightly coupled to a specific frontier model from a single provider now carries a government-approval window as part of its availability risk. Multi-provider design was already worth maintaining for cost reasons; the reason to keep it has broadened. Gemini 3.5 Pro missed its June target and is expected in July. When it lands, EO 14409 is the same framework in place.
The pricing for Sol, Terra, and Luna is real. Plan against it. Test on day one of GA. Keep your current baseline on what is actually accessible until then.
The updated GPT-5.6 pricing row is in /compare. For a breakdown of how Sol, Terra, and Luna stack up against Opus 4.8 and Sonnet 4.6, the model comparison table has the current figures.
Get the next post when it ships
One email on Sunday with the new post and a short list of what shipped that week — new guides, tool updates, and a couple of links worth reading.