OpenAI limits GPT-5.6 (Sol, Terra, Luna) to select government-approved partners; broader rollout in coming weeks
OpenAI announced a limited preview of its GPT-5.6 series on June 26: Sol (flagship for complex reasoning/coding), Terra (balanced mid-tier at 2x cheaper than Sol), and Luna (fastest, lowest-cost option). At government request, access is initially limited to approximately 20 trusted organizations whose participation has been shared with the Trump administration. OpenAI stated it plans to make all three models generally available in ChatGPT, Codex, and the API in the coming weeks.
OpenAI announced pricing: Sol is $5 input/$30 output per million tokens (same as GPT-5.5); Terra is $2.50/$15 (half Sol's cost); Luna is $1/$6 (most affordable). The company claims Sol delivers meaningful improvements in coding, cybersecurity, and agentic tasks with extended reasoning time and a new 'ultra mode' featuring subagents for complex projects. On its Preparedness Framework, OpenAI rates all three as High in cybersecurity and biological/chemical risk but below the Critical threshold, contrasting with Anthropic's stricter export approach.
In its announcement, OpenAI explicitly pushed back against the government approval process, writing: 'We don't believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.' The company framed this as a short-term step to enable broader availability while it works with the administration on a repeatable framework and new executive order language.
For architects: OpenAI is signaling that it will cooperate with government review but resist permanent approval gates, positioning itself as the willing-but-independent player between Anthropic (which resisted and got blocked harder) and the open-source labs (which released without restriction). Model tiering by capability and price is a new pattern: Sol for frontier, Terra for volume, Luna for scale. Teams should evaluate whether restricted partner access delays affect integration timelines; early pilot programs with approved partners may be the fastest path to production.
Sources
- Primary source
- techcrunch.com
“OpenAI is limiting the release of its newest AI models to a 'small group of trusted partners' at the behest of the U.S. government”
- help.openai.com
“GPT-5.6 Sol, our flagship and most capable model, alongside GPT-5.6 Terra, a strong lower-cost option, and GPT-5.6 Luna, our fastest and most cost-efficient model”
- openai.com
“We don't believe this kind of government access process should become the long-term default”