OpenAI has released GPT-5.6, introducing a complex model stratification system with 36 API variants and launching the ChatGPT Work and Codex superapp. The rollout included Luna, Terra, and Sol tiers alongside multiple effort levels, though the launch was marred by UX regressions that prompted OpenAI to reset usage limits.
- GPT-5.6 offers 36 configuration variants for API users, categorized into Luna, Terra, and Sol tiers with distinct effort settings.
- The ChatGPT interface now splits functionality into ChatGPT Work and Codex, removing the previous model picker in favor of a single slider for most users.
- OpenAI staff clarified that 'Max' extends time on hard problems while 'Ultra' parallelizes work across subagents, noting effort settings are not directly comparable to GPT 5.5.
- The launch experienced significant UX issues, including confusing project navigation and faster-than-expected usage depletion, leading to public course-correction and limit resets.
The release marks a shift toward explicit model selection and agent-based workflows, though the complexity of the new configuration options has drawn mixed reactions from the developer community.