OpenAI Unveils GPT-5.6 Trio: Sol, Terra, Luna Set New AI Benchmarks

OpenAI introduces a groundbreaking trio of AI models—Sol, Terra, and Luna—under the GPT-5.6 banner, setting new benchmarks in artificial intelligence. The preview, announced on June 26, 2026, marks a significant shift in the company's model architecture, with each tier tailored for specific use cases.

Sol, the flagship model, excels in complex problem-solving, agentic security, and long-horizon planning. It achieves a 91.9% score on Terminal-Bench 2.1, surpassing all competitors. Terra, the balanced everyday model, offers GPT-5.5-competitive performance at half the cost. Luna, the budget-friendly option, is designed for summarization, drafting, and routine automation.

Key Performance Metrics

The new models outperform their predecessors and competitors across various benchmarks. On Agent's Last Exam, Sol in code mode becomes the first model to cross 50% task completion, achieving 50.9%. In SecureBio, both Sol and Terra outshine GPT-5.5 while using fewer output tokens. On ExploitBench, all three GPT-5.6 models meet OpenAI's 'High' cybersecurity threshold.

Pricing and Market Strategy

OpenAI's pricing strategy reflects a competitive market approach. Sol is priced at $5 input and $30 output per million tokens, identical to GPT-5.5. Terra, at $2.50 input and $15 output per million tokens, competes directly with Anthropic's Claude Sonnet 4.6 and Google's Gemini 3.5 Flash. Luna, at $1 input and $6 output per million tokens, targets the ultra-budget segment, competing with DeepSeek Flash and Gemini Flash Lite.

Government-Gated Access

Access to the new models is currently restricted to approximately 20 vetted partner organizations via the API and Codex, as requested by the US government. OpenAI plans to broaden access in the coming weeks and emphasizes its commitment to broad accessibility. The company opposes making government-gated launches the norm, stating it 'believes in broad access.'

Industry Impact and Future Outlook

The simultaneous announcements of GPT-5.6's government-coordinated preview and the partial lifting of the export control ban on Claude Mythos 5 establish a new pattern for frontier AI releases in the US. This development signals a more regulated and controlled rollout of advanced AI technologies, balancing innovation with security concerns.

References

• AI News Today June 29 2026: 15 Biggest Stories