OpenAI Releases GPT-5.6 to the Public: Sol, Terra and Luna Explained

OpenAI has broadly released GPT-5.6, its newest frontier model family, roughly two weeks after limiting the rollout to a "small group of trusted partners" at the request of the U.S. government. The public release landed Thursday, July 9.

Three tiers, one new naming system

GPT-5.6 introduces a naming scheme OpenAI says will persist across generations: the number identifies the model's generation, while three durable capability tiers can advance on their own cadence:

• Sol — the flagship tier: $5 input / $30 output per million tokens
• Terra — the mid tier: $2.50 input / $15 output per million tokens
• Luna — the fast, low-cost tier: $1 input / $6 output per million tokens

The pitch is clearer choices across intelligence, speed, and cost — and notably, OpenAI is leaning on performance per dollar as its headline enterprise argument rather than raw capability alone.

The government-shaped rollout

The staggered release is the story inside the story. OpenAI held GPT-5.6 back from general availability at the U.S. government's request — echoing the export-control drama that forced Anthropic's Claude Fable 5 offline for several days in June. Frontier model releases now visibly run through a government review lane that didn't exist a year ago, and some analysts noted the public release timing landed almost exactly as restrictions on its chief rival expired.

GPT-Live: full-duplex voice

Alongside GPT-5.6, OpenAI announced GPT-Live, a new generation of voice models that can listen and speak at the same time. Full-duplex conversation removes the walkie-talkie turn-taking that has made voice assistants feel robotic — OpenAI says engaging with GPT-Live feels "much more like having a real conversation."

Why it matters

For consumers: the tier system means the free and low-cost ChatGPT experience gets a meaningful upgrade via Luna, while heavy users get Sol.

For builders: three price points on one generation simplifies model routing — and puts direct pressure on Anthropic's pricing, where Claude Fable 5 sits at $10/$50 per million tokens.

For decision-makers: the two-week government hold is the precedent to watch. Model access is now partly a policy question, and procurement plans should assume release schedules can slip on regulatory review.

What to watch

• Whether Sol tops independent benchmarks against Claude Fable 5 and Gemini's latest — early testing is still landing
• How fast GPT-Live reaches the API, and at what price
• Whether the "trusted partner" preview window becomes standard practice for every frontier release

Sources: OpenAI: GPT-5.6, CNBC, Engadget, Axios, OpenAI: GPT-Live