Key Takeaways
- OpenAI introduced a restricted preview of its GPT-5.6 series, featuring three models: Sol, Terra, and Luna
- The model names echo Solana’s SOL token and the infamous Terra/Luna blockchain collapse
- OpenAI clarifies the naming scheme represents performance levels, not cryptocurrency connections
- Sol serves as the premium offering, Terra provides mid-tier performance, and Luna targets budget-conscious users
- Federal officials requested OpenAI restrict initial access pending new cybersecurity policy development
On Thursday, OpenAI revealed its newest generation of artificial intelligence models under the GPT-5.6 banner, introducing three distinct versions labeled Sol, Terra, and Luna. The naming convention immediately sparked discussion within cryptocurrency communities due to striking parallels with prominent blockchain initiatives.
Sol matches the trading symbol for Solana, currently ranking among the top cryptocurrencies by market capitalization. Meanwhile, Terra and Luna reference a blockchain network that spectacularly imploded in 2022, erasing approximately $60 billion in investor wealth.
OpenAI emphasized that the nomenclature carries no intentional cryptocurrency association. According to the organization, these names simply denote varying performance tiers within their model lineup.
Breaking Down Model Capabilities
Sol represents the premium tier, engineered to tackle the most computationally intensive challenges. Terra occupies the middle ground, delivering performance comparable to the earlier GPT-5.5 iteration while cutting costs in half. Luna functions as the budget-friendly alternative, optimized for rapid processing and economic efficiency.
The Sol model introduces enhanced “max” and “ultra” reasoning capabilities. The ultra configuration deploys multiple collaborative sub-agents to accelerate problem-solving for intricate tasks.
OpenAI reported that Sol achieves record-breaking results on Terminal-Bench 2.1, a specialized assessment measuring command-line programming proficiency. The model also demonstrated enhanced capabilities in biological research and cybersecurity applications.
Regarding security applications, OpenAI noted that Sol assists in vulnerability detection and remediation. However, the company confirmed it remains below the “Cyber Critical” designation within their safety protocols, indicating it cannot autonomously generate complete functional exploits.
Controlled Rollout and Security Protocols
This announcement does not constitute a broad public deployment. OpenAI characterized the release as a “limited preview” accessible exclusively to select trusted collaborators. The organization continues conducting comprehensive safety evaluations before expanding availability.
According to reports, White House officials requested OpenAI maintain restricted access while federal agencies finalize updated cybersecurity executive directives.
OpenAI dedicated more than 700,000 GPU computing hours to automated adversarial testing, probing potential model vulnerabilities before deployment. Additionally, human security specialists examined the systems for exploitation possibilities.
The company outlined a multi-layered defense strategy incorporating embedded model safeguards, instantaneous content filtering, and user account surveillance.
API access pricing is established at $5 for input and $30 for output per million tokens for Sol. Terra costs $2.50 input and $15 output. Luna’s pricing structure is $1 input and $6 output.
OpenAI further announced plans to deploy Sol on Cerebras infrastructure this July, enabling processing speeds reaching 750 tokens per second.
The organization indicated all three variants will become more widely accessible through ChatGPT and API channels within the upcoming weeks.





