# OpenAI Launches GPT-5.6 Sol, Terra, Luna in Limited U.S. Government Preview

> The tiered frontier models enter restricted access first to enable security-vetted evaluation before planned expansion to general availability in the coming weeks.

*Published 2026-06-28 · By Marcus Vance*

GPT-5.6 is OpenAI's tiered frontier model series consisting of Sol, Terra and Luna variants introduced through a limited preview.

OpenAI has begun a limited preview of the GPT-5.6 series to allow for security-vetted access as requested by the U.S. government. The preview includes the flagship Sol model along with Terra and Luna variants. Access is currently limited to a small group of trusted partners through the API and Codex. This strategy reflects a commitment to responsible deployment of advanced AI technologies.

The decision to start with restricted availability marks a notable change in release practices for frontier models. By engaging only vetted users initially, OpenAI can monitor performance and address any issues before wider distribution. The company has stated its intention to expand access broadly in the near term.

## What prompted the limited preview of the GPT-5.6 series?

The U.S. government requested the limited preview to ensure that the most advanced AI models receive appropriate security considerations from the outset. This request led OpenAI to restrict the initial rollout to trusted partners. The approach prioritizes safety and controlled evaluation over immediate broad market entry. Such governmental involvement highlights increasing regulatory scrutiny on frontier AI systems.

OpenAI's compliance with this request demonstrates alignment with national security priorities. The limited group allows for detailed feedback collection in a secure environment. This could serve as a model for future interactions between AI companies and government entities. The preview phase is expected to inform refinements ahead of public release.

## How do the Sol, Terra, and Luna models compare in capabilities and pricing?

Sol represents the top tier with the highest performance levels suited for demanding applications. It features an ultra mode that incorporates sub-agents to boost results on complex tasks. Terra is positioned as a balanced option that delivers competitive results to earlier models at reduced expense. Luna focuses on speed and affordability while maintaining solid capabilities for a wide range of uses.

The pricing reflects these differences with Sol at the premium end and Luna at the lowest cost structure. This tiering enables users to match model selection to their operational requirements and financial considerations. The structure supports scalability from individual developers to large organizations. OpenAI has designed the lineup to address diverse market segments effectively.

Comparison of GPT-5.6 VariantsVariantPositioningTerminal-Bench 2.1 ScoreInput Price per 1M TokensOutput Price per 1M TokensSolFlagship88.8% (91.9% Ultra)$5$30TerraBalancedCompetitive to GPT-5.5$2.50$15LunaAffordableStrong capability$1$6

The comparison table shows clear distinctions in both performance metrics and associated costs. Sol's higher scores justify its pricing for users requiring maximum capability. Terra and Luna provide value propositions that lower the barrier to entry for advanced AI tools. These options may drive increased utilization across industries.

## What benchmark performance has GPT-5.6 Sol demonstrated?

GPT-5.6 Sol has established new state of the art results on the Terminal-Bench 2.1 coding benchmark. The base score stands at 88.8 percent while the ultra mode with sub-agents reaches 91.9 percent. These figures indicate superior handling of coding tasks compared to prior models. The efficiency in token usage further enhances its practical value for developers.

The benchmark success underscores advancements in the model's reasoning and problem-solving abilities. Sub-agents in ultra mode allow for specialized handling of subtasks within larger projects. This could translate to improved productivity in software engineering workflows. Continued evaluation will reveal additional strengths in other domains.

> We're beginning a limited preview of the GPT‑5.6 series: Sol, our flagship model; Terra, a balanced model for everyday work; and Luna, a fast and affordable model. Terra has competitive performance to GPT‑5.5 while being 2x cheaper and Luna brings strong capability at our lowest cost.OpenAI

## What rollout timeline has OpenAI outlined for broader access?

Following the limited preview, OpenAI plans to make the models generally available within the coming weeks. The transition will extend access to a wider audience through standard channels. This phased method allows time for optimization based on initial user experiences. The company maintains a focus on broad access while incorporating safety measures.

- Begin with limited preview for trusted partners in API and Codex
- Collect performance data and feedback during preview phase
- Prepare for general availability rollout
- Enable ongoing independent advancements for each tier

The ordered list above outlines the planned sequence of events leading to full public access. Each step builds upon the previous to ensure a smooth and secure expansion. This methodical process aligns with the emphasis on responsible AI development. Stakeholders can anticipate progressive increases in availability.

## What are the market implications of the GPT-5.6 tiered launch?

The tiered pricing and performance options are likely to influence competitive dynamics in the frontier AI space. Lower costs for Terra and Luna could attract new users who previously found advanced models prohibitively expensive. Government involvement in the preview may encourage other developers to adopt similar safety protocols. Overall, the launch supports greater democratization of AI capabilities through accessible pricing.

Enterprises stand to benefit from the ability to select appropriate models for specific workloads. This flexibility can lead to more efficient allocation of computational resources. The naming convention with durable tiers facilitates long-term planning and updates. Market observers will watch how adoption rates evolve post-general availability.

## What comes next for the GPT-5.6 series and OpenAI?

The coming weeks will see the expansion to general availability as announced. Each model tier is positioned to receive independent updates as development progresses. This allows OpenAI to respond to user needs and technological opportunities in a targeted manner. The series represents a step toward more nuanced AI product strategies.

Future iterations may build on the benchmark achievements and cost efficiencies established here. The focus on safety-vetted access could become a standard practice for subsequent releases. OpenAI continues to navigate the balance between innovation speed and responsible deployment. The outcomes of this preview will shape those future directions.

## Sources

1. [GPT-5.6 Sol sets a new state of the art on Terminal-Bench 2.1](https://openai.com/index/previewing-gpt-5-6-sol/)
2. [OpenAI is beginning a limited preview of GPT-5.6 Sol, our flagship and most capable model, alongside GPT-5.6 Terra... We plan to make GPT-5.6 Sol, Terra, and Luna generally available in the coming weeks.](https://help.openai.com/en/articles/20001325-a-preview-of-gpt-56-sol-terra-and-luna)
3. [We believe in broad access and plan to make GPT-5.6 Sol, Terra, and Luna generally available in the coming weeks. For now, at the request of the U.S. government, we’re starting with a limited preview among a small group of trusted partners in Codex and the API.](https://x.com/OpenAI/status/2070555273467687257)

---
Source: https://aiintelreport.com/frontier-models/openai-gpt-5-6-limited-preview
Index: https://aiintelreport.com/llms.txt · Full text: https://aiintelreport.com/llms-full.txt