OpenAI Unveils GPT-5.6 Models in Limited Preview, Citing U.S. Government Coordination

OpenAI Unveils GPT-5.6 Family: Sol, Terra, Luna

OpenAI has launched a limited preview of its next-generation GPT-5.6 model series, introducing three distinct, capability-tiered models: GPT-5.6 Sol, GPT-5.6 Terra, and GPT-5.6 Luna. The announcement was made on June 26, 2026, following the release of the GPT-5.6 Preview System Card on June 25. This new family arrives two months after the GPT-5.5 release, marking a significant advancement in OpenAI’s offerings.

The naming convention for this series introduces a new approach, where the number "5.6" identifies the model’s generation. Sol, Terra, and Luna, inspired by the Sun, Earth, and Moon, denote durable capability tiers designed to evolve independently. This system aims to provide developers and users with clearer choices based on intelligence, speed, and cost requirements.

Advanced Capabilities and Tiered Pricing

GPT-5.6 Sol stands as the flagship and most capable frontier model, engineered for re-engineering developer and enterprise workflows. OpenAI highlights its major performance gains for long-running coding, cybersecurity, and agentic tasks. Sol also advances the frontier in software engineering, computer use, professional knowledge work, scientific research, and biology.

GPT-5.6 Terra is positioned as a balanced model for efficient, everyday work, offering a strong lower-cost option. For high-volume tasks requiring speed and cost-efficiency, GPT-5.6 Luna is introduced as the fastest and most cost-effective model in the family.

The pricing structure for the GPT-5.6 models is set per 1 million tokens. Sol is priced at $5 for input tokens and $30 for output tokens, matching the pricing of GPT-5.5. Terra costs $2.50 per million input tokens and $15 per million output tokens, while Luna is the most affordable at $1 per million input tokens and $6 per million output tokens.

A New Architectural Approach to AI Reasoning

A core architectural evolution within the GPT-5.6 series, particularly for Sol, centers on how compute resources are allocated during inference. OpenAI is introducing a new "max" reasoning effort mode, which explicitly grants the flagship Sol model extended time to reason deeply through highly complex problems. Beyond this, an "ultra" mode makes its debut, expanding past the structural boundaries of a single standalone model. This mode deploys specialized "subagents" to divide, conquer, and accelerate multi-step, long-horizon projects. Initial evaluations suggest that this subagent coordination significantly shifts the frontier for programmatic execution.

OpenAI emphasizes that the GPT-5.6 Sol, Terra, and Luna models have been developed with its most robust safeguards to date, with configurations specifically matched to each model’s capabilities.

Phased Rollout Amidst Government Coordination

The initial rollout of the GPT-5.6 series is available through the OpenAI API and Codex to a narrow and limited group of trusted partners and organizations. During this preview phase, the models are not accessible via ChatGPT. OpenAI plans to make the family generally available across ChatGPT, Codex, and the API in the coming weeks. The preview is not a broad self-service program, and there is no public application or waitlist for participation; access is subject to review and enabled on a rolling basis, scoped to specific API organizations and Codex workspaces.

This phased release marks a highly unusual chapter in AI deployment, attributed to ongoing coordination with the U.S. government. OpenAI previewed its plans and the models’ capabilities to the government ahead of the launch, following an executive order issued by President Donald J. Trump on June 2, 2026. This order calls upon various federal agencies to collaborate on a process for benchmarking and assessing new AI models to ensure their safety and appropriateness for wide release. At the government’s request, OpenAI is starting with a limited preview for trusted partners whose participation has been shared with the government, allowing for additional testing and coordination while safety and usability continue to be improved. Enterprise buyers during this period must navigate a novel landscape of real-time safety interventions, mandatory compliance parameters, and structured token caching systems.

The coordinated, phased release of OpenAI's new models, particularly under government request, signals a significant shift in AI deployment, potentially setting a precedent for future frontier model rollouts.