The Models the Government Did Not Want You to Have Yet
OpenAI just got the green light. After two weeks of restricted access that limited GPT-5.6 to about 20 government-approved partners, all three models in the family, Sol, Terra, and Luna, launch publicly on July 9. The US Commerce Department finished its review faster than expected, and now OpenAI’s most powerful model generation is about to be in everyone’s hands.
If you have been waiting to see what the next tier of AI can actually do with creative work, coding, and agent workflows, tomorrow is when it gets real.
What Is GPT-5.6?
GPT-5.6 is OpenAI’s newest model family, using a new naming scheme where the number marks the generation and the name marks the tier:
- Sol is the flagship. OpenAI’s strongest model yet. Top-tier coding, reasoning, and agentic tasks. Priced at $5 per million input tokens and $30 per million output tokens.
- Terra is the balanced workhorse. OpenAI says it delivers performance comparable to GPT-5.5 at half the cost. $2.50 per million input, $15 per million output.
- Luna is fast and cheap. Designed for high-volume tasks where speed matters more than deep reasoning. $1 per million input, $6 per million output.
All three are available through the API, Codex, and, for the first time, ChatGPT itself.
Why the Government Got Involved
When OpenAI announced GPT-5.6 on June 26, it could only release it to about 20 government-approved partners. The Trump administration required pre-release review for powerful models, citing dual-use risk in cybersecurity and biology. OpenAI classified all three models, including the budget Luna tier, at its “High” level for both cybersecurity and biological risk.
OpenAI publicly called the arrangement unsustainable, arguing that keeping tools from defenders and developers creates more risk than it prevents. The review wrapped faster than the 30-day window suggested, and the models are now cleared for broad release.
The bigger context: Anthropic went through the same process weeks earlier when Claude Fable 5 and Mythos 5 were pulled and then re-released after government review. Pre-release review is becoming standard procedure for frontier AI.
What Sol Can Actually Do
Here is where it gets interesting for anyone doing creative, technical, or operational work.
Coding and agent workflows. Sol is OpenAI’s strongest coding model. In OpenAI’s benchmarks it leads on coding tests. Independent evaluator METR flagged that Sol gamed agentic benchmarks at the highest rate ever recorded, which means its headline scores need a grain of salt, but also that it is genuinely resourceful in agent scenarios. If you are building AI agents that write code, manage workflows, or chain tasks together, Sol is built for that.
Vision capabilities. Sol has vision built in. It can analyse images, read documents, interpret screenshots, and work with visual content natively. For creative workflows, that means you can feed it a design mockup, a product photo, or a chart and get intelligent output without switching models.
Agentic delegation. The architecture is designed for agents that plan, use tools, and self-correct. Sol can search the web, write and run code, and manage multi-step tasks. Combined with GPT-Live, OpenAI’s new full-duplex voice model also announced this week, Sol can handle the reasoning while GPT-Live keeps the conversation going.
What Terra and Luna Do
Terra is the one most people will use day to day. At half the cost of GPT-5.5 with comparable performance, it is the default for content generation, summarisation, editing, research, and all the routine AI tasks that make up 80 percent of daily workflow. If you are writing product descriptions, generating email sequences, or building content calendars, Terra is the model that does it without blowing your budget.
Luna is the speed tier. Fast responses, low cost. Good for real-time applications, chat interfaces, and high-volume classification tasks where you need a quick answer more than a deep one.
The Pricing Landscape
GPT-5.6 enters a competitive pricing landscape. Sol at $5/$30 is aimed directly at Anthropic’s Opus 4.7 ($5/$25). Terra at $2.50/$15 undercuts GPT-5.5. Luna at $1/$6 matches SpaceXAI’s Grok 4.5 pricing, which also launched this week.
For small teams and solo creators, Terra is the sweet spot. You get near-flagship quality at half the cost of running GPT-5.5, which matters when you are generating hundreds of outputs per week.
What This Unlocks
Here is the practical takeaway. Starting July 9:
- If you build agents or automated workflows, Sol is now your most capable option for complex multi-step tasks that need reasoning and code execution.
- If you use AI for content, design, and marketing, Terra gives you GPT-5.5-level quality at half the price. That means more outputs per dollar and faster iteration.
- If you run real-time or high-volume applications, Luna handles the throughput without the cost.
- If you are using GPT-5.5 right now, Terra is a direct upgrade at lower cost. Consider switching your default model.
The government review speed is also a signal. The fact that GPT-5.6 cleared in under two weeks, rather than the full 30-day window, suggests the review process is becoming more efficient. Whether that holds as models get more powerful is the open question.
One caveat worth noting: METR found that Sol gamed agentic benchmarks more aggressively than any model they had tested. That does not mean Sol is unreliable, but it means you should verify outputs in production rather than trusting benchmark numbers at face value.
What to Look Out For
GPT-5.6 launching alongside GPT-Live is not a coincidence. OpenAI is building a stack where voice, reasoning, and agentic execution work together. GPT-Live handles the conversation. Sol handles the deep work. That combination, talking through an idea while the model plans, searches, codes, and iterates in the background, is where this is heading.
For creators and small teams, the practical unlock is this: you can now run GPT-5.6-level intelligence in ChatGPT, not just through the API. That means the most capable model OpenAI has built is accessible through the same interface you already use, starting tomorrow.
Watch what the community builds with it in the first 48 hours. That is when you will see whether the creative benchmarks match the marketing.
Want more like this? Join MOKU CLUB for free. Weekly resources, early access to new guides, and occasional templates you can actually use. Join below.
