GPT-5.6 was released publicly on July 9, 2026, as a family of models that comes in three distinct variants, ranked from least to most capable: Luna, Terra, and Sol. OpenAI described the launch, previewed in its introducing GPT-5.6 Sol announcement, as bringing its new flagship, Sol, alongside Terra, a balanced model for everyday work, and Luna, its most cost-efficient model, with Sol achieving state-of-the-art results across coding, knowledge work, cybersecurity, and science. The public rollout came roughly two weeks after the company limited access to a small group of trusted partners at the request of the U.S. government. The restriction marked the first time the U.S. government had preemptively asked an American AI company to restrict the launch of a model before release.
Three tiers, one generation
In the new naming system introduced with GPT-5.6, the number identifies a model’s generation, while Sol, Terra, and Luna identify durable capability tiers that can advance on their own cadence. The naming scheme replaces the company’s earlier “mini” and “nano” suffixes.
The flagship model, Sol, is considered its workhorse, suited for complex agentic capabilities in coding, biology, and cybersecurity. Terra is a balanced model for everyday work that performs similarly to GPT-5.5 at half the cost. Luna is a fast and affordable model that brings strong capability at the lowest cost offered by the company.
The family introduces Programmatic Tool Calling and a Sol Ultra mode that uses parallel sub-agents for demanding tasks. GPT-5.6 also introduces more predictable prompt caching, including support for explicit cache breakpoints and a 30-minute minimum cache life.
On Agents’ Last Exam, Sol records 53.6, beating Claude Fable 5 by 13.1 points. On SWE-Bench Pro, however, Sol scores 64.6% compared to Claude Fable 5’s 80%, a gap of 15.4 points.
A government review first
OpenAI announced three new artificial intelligence models and said it was complying with the U.S. government’s request to initially limit the rollout to a “small group of trusted partners.” On June 2, 2026, President Trump signed an executive order titled “Promoting Advanced Artificial Intelligence Innovation and Security,” directing federal agencies to establish a framework for the secure deployment of frontier AI models, including a process by which developers would voluntarily provide the government with early access to models for up to 30 days before releasing the technology to other trusted partners.
The U.S. Department of Commerce gave its approval to OpenAI for a broad launch of the GPT-5.6 series after additional testing and meetings with government agencies. The limited preview had begun on June 26, 2026; on the evening of July 8, OpenAI announced on X that all three models would launch publicly the next day. That 13-day restricted period drew scrutiny from across the industry. The OpenAI situation was notable because the company was complying with a government request rather than following a legally binding order.
OpenAI stated it was working with government partners to assess the safety of the models prior to their release, an approach the company does not believe “should become the long-term default.”
Pricing and developer access
GPT-5.6 is priced per million tokens: Sol is $5 input and $30 output; Terra is $2.50 input and $15 output; and Luna is $1 input and $6 output.
GPT-5.6 introduces more predictable caching with explicit cache breakpoints and a 30-minute minimum cache life.
The pricing spread is significant: Luna’s output cost is one-fifth of Sol’s, giving developers a sharper tool-routing decision than any prior OpenAI family offered.
GPT-5.6 Sol, Terra, and Luna are generally available across ChatGPT, Codex, and the OpenAI API. Sol supports a 1.05 million-token context window with 128,000 tokens of maximum output. Alongside the models, OpenAI also launched ChatGPT Work, a new mode for complex tasks.
What it means for the market
The public launch of the GPT-5.6 models came after Anthropic saw its Fable 5 and Mythos 5 models pulled from foreign access in June under government direction, before controls were lifted and access was restored on July 1. The rollout landed in a competitive week as multiple frontier labs moved aggressively on pricing and capability.
For enterprises already tracking the rapid growth of AI productivity tools, the three-tier structure signals a broader industry shift toward model families priced by use case rather than by raw capability alone.
As observers tracking the evolving AI assistant market have noted, the competition is increasingly about price, speed, and fit for specific workflows rather than benchmark supremacy. The defining shift of July 2026, across the whole model landscape, may be the move from “best model wins” to “best fit wins.”
