OpenAI's GPT-5.6 Sol, Terra, Luna launch in limited partner preview
Three new models focused on agentic coding and safety, but not for ChatGPT users yet.
OpenAI has begun rolling out its GPT-5.6 family of models — named Sol, Terra, and Luna — in a limited preview to selected API and Codex partners. The preview, effective July 7, 2026, does not include public availability through ChatGPT. Sol serves as the flagship tier, designed for high-value, complex tasks such as agentic coding, professional knowledge work, scientific research, and cybersecurity workflows. Terra and Luna likely target more cost-sensitive or specialized use cases.
The GPT-5.6 family emphasizes a more robust safety stack alongside stronger capabilities. By restricting early access to partners, OpenAI can gather feedback and validate safety measures before a wider release. This strategy mirrors previous launches and signals a continued focus on responsible deployment. Professionals in research, security, and software engineering should watch for future public availability.
- Three models: Sol (flagship), Terra, and Luna, tailored for different task tiers.
- Initial access limited to API and Codex partners; no ChatGPT integration yet.
- Focus areas: agentic coding, knowledge work, scientific research, and cybersecurity with enhanced safety.
Why It Matters
OpenAI's tiered preview approach gives pros early access to cutting-edge AI for specialized workflows while ensuring safety.