OpenAI is changing the game again. Just weeks after its last major update, the company has launched GPT-5.6—not as a single model, but as a trio of specialized AI tools released under a cautious new government-cooperative strategy.
The launch, which took place on June 26, 2026, introduces a new family of models and a phased rollout developed in partnership with the U.S. government.
A New Family of Models: Sol, Terra, and Luna
Instead of a single successor, GPT-5.6 arrives as a suite of three models, each designed for different use cases and named after celestial bodies:
- Sol: The flagship and most powerful model.
- Terra: A balanced model intended for everyday tasks.
- Luna: A high-speed, cost-effective model.
This new naming system signifies a major shift in OpenAI's strategy. The number "5.6" identifies the model's generation, while "Sol, Terra, and Luna" represent distinct capability tiers.
This structure aims to give developers and users clearer choices to balance intelligence, speed, and cost for their specific needs.
max reasoning effort "to give Sol the most time to reason deeply" and an ultra mode that uses subagents to tackle complex work more efficiently.Phased Release and Government Cooperation
Initially, GPT-5.6 launched in a limited preview available only to "a select group of trust partners and organizations." This staggered release was a direct result of cooperation with the U.S. government, following a recent AI executive order signed by President Donald Trump.
The order requires AI developers to voluntarily provide cutting-edge models to the government for assessment before a full public release. OpenAI complied, sharing its list of initial participants directly with the government.
More recently, OpenAI announced it will proceed with a wider public release of all three models. The move follows a similar situation with rival Anthropic, which restored access to its latest models after complying with a temporary export control directive from the Department of Commerce.
Performance, Pricing, and New Features
With the regulatory hurdles cleared for a wider launch, OpenAI has also revealed how developers will access and pay for this new power. The new model family comes with a tiered pricing structure based on usage per one million tokens:
| Model | Input Cost | Output Cost |
|---|---|---|
| Sol | $5.00 | $30.00 |
| Terra | $2.50 | $15.00 |
| Luna | $1.00 | $6.00 |
Alongside the new models, OpenAI is introducing a more predictable prompt caching system with a 30-minute minimum cache life.
For GPT-5.6 and future models, cache writes will be billed at 1.25x the standard input rate. Reads from the cache, however, will receive a 90% discount.
For high-performance applications, OpenAI is also launching GPT-5.6 Sol on the Cerebras platform this month, promising speeds of up to 750 tokens per second. Access will initially be limited as capacity expands.
Cybersecurity and Internal Use
Beyond public use, OpenAI is already putting GPT-5.6 to work on two critical fronts: cybersecurity and its own internal research. The company claims the models were developed with its "most robust safeguards to date" and are being deployed for defensive cyber tasks.
These tasks include secure code review, threat modeling, and malware analysis through its Trusted Access for Cyber program.
The model is also dramatically accelerating OpenAI's own research. During internal testing, the average daily output tokens per active researcher more than doubled compared to the highest levels seen with GPT-5.5.
Coinciding with the GPT-5.6 announcement, OpenAI rolled out a new generation of voice models called GPT-Live. These models, GPT-Live-1 and GPT-Live-1 mini, can listen and speak simultaneously, enabling more natural conversations for ChatGPT users worldwide.
What This Means for the AI Race
The launch of GPT-5.6 signals a new chapter for OpenAI and the broader AI industry. The shift from a single flagship model to a tiered family offers users unprecedented flexibility, but it also reflects a market that is maturing beyond a one-size-fits-all approach.
Furthermore, the close cooperation with government regulators establishes a new precedent for how frontier models are introduced to the public. This cautious, safety-first strategy could become the standard as AI capabilities continue to accelerate. For developers and competitors, the message is clear: the race is no longer just about raw power, but about specialization, safety, and strategic deployment.