OpenAI released GPT-5.6 on June 26, 2026, and it arrived split into three separate models. Sol is the flagship built for the hardest problems, Terra is the balanced workhorse, and Luna is the fast, low-cost option for everyday tasks. For the first time, a major American AI model launched behind a government-managed access list, available only to roughly 20 approved organizations through OpenAI’s API and Codex.
That gating is the real story here. GPT-5.6 is not in ChatGPT yet, and most people cannot touch it. This article breaks down what each of the three tiers actually does, what they cost, the new “ultra” mode that splits work across sub-agents, and the timeline for when GPT-5.6 reaches the rest of us. We also explain why Washington put a frontier model on a short leash, and how it connects to the wave of release delays already hitting Anthropic.
The Key Takeaways
- GPT-5.6 launched June 26, 2026 as three models, Sol (most powerful), Terra (balanced), and Luna (fast and cheap).
- Access is restricted to about 20 US-government-approved organizations via the API and Codex; broad release in ChatGPT is “coming weeks.”
- It is the first US frontier model to ship under a government-managed access list, a step beyond the voluntary review set up by the June 2 AI executive order.
- A new “ultra” mode deploys sub-agents to break complex tasks apart, alongside a new “max” reasoning setting.
- Reported API pricing per 1M tokens runs $5/$30 (Sol), $2.50/$15 (Terra), and $1/$6 (Luna).
What Is GPT-5.6?
GPT-5.6 is OpenAI’s newest model family, and it is the company’s biggest step forward in reasoning and agentic work since GPT-5.5. Instead of one model, OpenAI shipped a tiered lineup so you can pick the right balance of power, speed, and cost for the job. In its official GPT-5.6 preview, OpenAI says the family advances the frontier on software engineering, computer use, professional knowledge work, scientific research, and cybersecurity.
The naming follows a sun-and-moon theme. Sol is the bright, heavyweight model for the toughest problems, Terra sits in the middle for high-volume business work, and Luna is the lightweight option tuned for speed and low cost. All three are previewing through the OpenAI API and Codex first, with ChatGPT and wider API access following later.
Sol, Terra, and Luna: The Three GPT-5.6 Tiers
Each GPT-5.6 model targets a different kind of work, which makes the lineup easier to reason about than a single do-everything model. Here is how OpenAI frames the split.
| Model | Best for | Input (per 1M) | Output (per 1M) |
|---|---|---|---|
| GPT-5.6 Sol | The hardest problems, complex coding, security research | $5 | $30 |
| GPT-5.6 Terra | High-volume business tasks, customer support, internal tools, document analysis | $2.50 | $15 |
| GPT-5.6 Luna | Fast everyday work, summarization, drafting, routine automation | $1 | $6 |
Pricing reflects figures reported at launch by VentureBeat and may shift before general availability.
GPT-5.6 Sol
Sol is the flagship, the model you reach for when accuracy matters more than speed or budget. OpenAI positions it for the hardest problems, think deep coding sessions, complex agentic work, and security research. It is also the most expensive tier, so it earns its place on the jobs a lighter model would fumble.
GPT-5.6 Terra
Terra is the balanced middle, built to run at scale without burning through tokens. It targets high-volume business work like customer support, internal tools, and document analysis, where you need solid reasoning on a predictable budget. For most production workloads, Terra is the sensible default.
GPT-5.6 Luna
Luna is the cheapest and fastest of the three, sitting close to GPT-5.5 on several tests while costing a fraction to run. OpenAI aims it at everyday work like summarization, drafting, and routine automation. When volume matters more than raw power, Luna keeps the bill low.
GPT-5.6 Pricing
The pricing spread is wide on purpose. Luna at $1 input and $6 output per million tokens is roughly six times cheaper on output than Sol at $5 and $30, so you only pay flagship rates when the task truly needs it. Terra lands in between at $2.50 and $15, giving you most of the capability without the top-tier bill.
For high-volume jobs like classification, summarization, or first-draft writing, Luna keeps costs low. For agentic coding and security research, Sol is the one worth paying for. This tiering mirrors how the rest of the market is moving, and you can see the full field on our roundup of the best AI models right now.
The New “Ultra” Mode and Sub-Agents
The most interesting upgrade in GPT-5.6 is not raw benchmark scores, it is how the model thinks. OpenAI added a new “max” reasoning effort setting for harder problems, plus an “ultra” mode that deploys sub-agents to decompose a complex task into smaller pieces and work them in parallel.
That approach echoes a pattern showing up across the industry, where one model coordinates a pool of workers rather than grinding through everything in a single pass. It points toward longer, more autonomous tasks, and OpenAI reports meaningful gains on multi-hour Codex computer-use workloads compared with GPT-5.5. A reported 1.5 million token context window and a roughly May 2026 knowledge cutoff also circulated at launch, though neither is confirmed in OpenAI’s official documentation yet, so treat both as unverified for now.
Why You Can’t Use GPT-5.6 Yet
Here is the part that makes GPT-5.6 different from every previous OpenAI launch. The models went live only to about 20 organizations that the US government approved, after OpenAI shared the models and its release plans with Washington. This is the first time an American AI company has launched a frontier model under a government-managed access list, a real step beyond the US government’s new model-review process that the June 2 AI executive order set up as voluntary.
OpenAI was not thrilled about it. As reported by TechCrunch, the company limited the rollout after a government request and made clear it disagrees with the precedent, stating it does not believe this kind of government access process should become the long-term default. The move follows the same playbook that hit Anthropic, where Claude Fable 5 was pulled offline by a US export-control order just days after launch. Frontier releases are now running through Washington first.
When Will GPT-5.6 Be Generally Available?
OpenAI says it expects to expand access to more companies within a week of launch, with the goal of a broad release in ChatGPT, Codex, and the API in the coming weeks. There is no firm public date, and the timeline depends partly on how the government review process plays out.
For now, GPT-5.6 sits behind a trusted-partner wall. If you are not on the access list, the practical answer is to wait, watch the rollout, and keep using GPT-5.5 and the other current flagships in the meantime.
How GPT-5.6 Fits the Bigger Picture
GPT-5.6 lands in the middle of an intense stretch for frontier AI. Anthropic is navigating its own government-driven delays, Google’s Gemini 3.5 Pro slipped toward July, and the competitive gap at the top keeps narrowing. If you want the context on who is actually ahead, our breakdown of the Anthropic vs OpenAI race covers how the two labs stack up in 2026.
When GPT-5.6 does open up, the smart play is not locking yourself to one provider. With Fello AI, you reach OpenAI’s models alongside Claude and Gemini in a single app on your Mac and iPhone, so you can switch to whichever model is strongest or actually available on any given day. As government gating makes single-vendor access less predictable, having every major model in one place stops a restricted rollout from blocking your work.
Conclusion
GPT-5.6 is a genuine leap, three sharply differentiated tiers, a new sub-agent ultra mode, and stronger agentic performance. The catch is access, since it launched to roughly 20 approved organizations under the first US government-managed access list for a frontier model. Watch for the broad ChatGPT and API rollout in the coming weeks, and in the meantime keep your options open across providers so a gated launch never stalls your workflow.
FAQ
What is GPT-5.6?
GPT-5.6 is OpenAI’s newest model family, released June 26, 2026, in three tiers, Sol for the hardest problems, Terra for high-volume business work, and Luna for fast, low-cost everyday tasks. It advances reasoning, coding, and agentic performance over GPT-5.5.
What are Sol, Terra, and Luna?
They are the three GPT-5.6 models. Sol is the most powerful and most expensive, Terra is the balanced mid-tier, and Luna is the fastest and cheapest. You pick the tier that matches the task’s difficulty and your budget.
Why is GPT-5.6 restricted?
OpenAI limited GPT-5.6 to about 20 US-government-approved organizations after sharing the models with Washington. It is the first US frontier model to launch under a government-managed access list, going beyond the voluntary review the June 2 AI executive order created.
When can I use GPT-5.6 in ChatGPT?
OpenAI says a broad release in ChatGPT, Codex, and the API is coming in the weeks after launch, with access expanding to more companies first. There is no confirmed public date yet.
How much does GPT-5.6 cost?
Reported API pricing per 1M tokens is $5 input and $30 output for Sol, $2.50 and $15 for Terra, and $1 and $6 for Luna. These figures may change before general availability.
