OpenAI Releases GPT-5.6: Sol, Terra and Luna Variants

Announcement Details

OpenAI unveiled the GPT-5.6 family on July 9, 2026. The release consists of three variants named Sol, Terra, and Luna. Sol serves as the main model for demanding tasks. Terra targets intermediate workloads. Luna provides a lower-cost option. The models are positioned for enterprise operations, coding assistance, and scientific research. They are now accessible through ChatGPT, Codex, and the OpenAI API.

Coding Performance Metrics

OpenAI states that Sol reaches 80 on the Artificial Analysis Coding Agent Index. This score sits 2.8 points above Anthropic's Fable 5. The model uses less than half the output tokens, runs in less than half the time, and costs roughly one-third less. Terra exceeds Fable 5 on the same index. Luna surpasses Opus 4.8. These figures focus on token efficiency during code generation and agent-style tasks. Developers working with long context windows may see reduced API spend if the reported ratios hold across typical codebases. Integration remains standard through the existing OpenAI client libraries with no new endpoints required.

Cybersecurity Capabilities

The GPT-5.6 family is described as the strongest cybersecurity model from OpenAI to date. It supports threat modeling, automated code review, vulnerability patching, and blue team exercises. Performance claims emphasize frontier results achieved with fewer tokens than prior releases. The Trump administration had considered restrictions on rollout due to potential misuse risks. Production use for offensive security tooling would still require separate safeguards and compliance checks. Teams running internal red team simulations can now route those workloads to Luna for cost control while reserving Sol for final validation steps.

ChatGPT Work and Deployment

OpenAI introduced ChatGPT Work as a desktop, web, and mobile companion for enterprise teams. The tool handles document drafting, spreadsheet updates, and presentation generation. It runs on the GPT-5.6 family and maintains the same token pricing structure as the core API. No additional infrastructure changes are needed beyond existing OpenAI account settings. Organizations already using Codex for code review can extend the same keys to ChatGPT Work without separate authentication flows. Trade-offs remain around data residency and the standard requirement to review all generated outputs before production use.

---

📖 Related articles

• Agentic Coding: Una Trappola per lo Sviluppo Software?
• File agents.md: utili per gli agenti di coding?
• Lean-ctx: Ottimizzatore Ibrido Riduce Consumo Token LLM del 89-99%