GPT-5.6 released with Sol, Terra, Luna, and ChatGPT Work
OpenAI lists GPT-5.6 as a released model family across ChatGPT, Work, Codex, and the API by plan.
Verify official source →GPT-5.6-powered agentic workspace for longer real work tasks.
OpenAI's agentic work mode in ChatGPT for longer tasks across apps, files, documents, spreadsheets, reports, Sites, Codex, and connectors.
Opens official website
Live change feed
Shared with the homepage and tool pages
OpenAI lists GPT-5.6 as a released model family across ChatGPT, Work, Codex, and the API by plan.
Verify official source →OpenAI's GPT-5.6 Help Center documentation lists Work in ChatGPT as a product surface where Sol, Terra, and Luna are available for eligible Plus, Pro, Business, and Enterprise users.
ChatGPT Work is OpenAI's GPT-5.6-powered agentic work surface for longer, multi-step tasks. Instead of only answering a prompt, Work is designed to operate across apps, files, documents, spreadsheets, reports, Sites, Codex, and connected tools. OpenAI's GPT-5.6 availability notes list Sol, Terra, and Luna for Work in ChatGPT on eligible Plus, Pro, Business, and Enterprise plans. For teams, the key buying question is not only model quality; it is governance: what the agent can read, which connectors it can use, whether approvals are required, how actions are logged, and how admins limit sensitive data access.
Longer Work Execution
Designed for multi-step tasks that span files, documents, spreadsheets, reports, Sites, Codex, and connected apps.
GPT-5.6 Model Family
Uses GPT-5.6 Sol, Terra, and Luna on eligible Work plans, giving teams a flagship, balanced, and fast option.
Enterprise Governance Requirement
The practical deployment question is how permissions, connectors, approvals, logging, and sensitive data boundaries are managed.
Prices, plan limits, and promotions may change. Verify final pricing on the official site. Last verified: 2026-07-10.
ChatGPT Work is strategically important because it makes ChatGPT a work-execution layer. It belongs in AI Agent Tools and business-agent rankings, but teams should evaluate it through governance and cost controls, not just model capability.
Model access status
GPT-5.6 · Released · risk Medium