GPT-5.6 Availability: ChatGPT Model Picker, Codex and API

OpenAI released GPT-5.6 for general availability on July 9, 2026, replacing the earlier limited-preview status. The GPT-5.6 family is now available across ChatGPT, Codex, and the OpenAI API, but the models and controls shown to a user still depend on plan, product, rollout state, and workspace settings.

GPT-5.6 rollout status

OpenAI says the global rollout began on July 9 and was scheduled to continue gradually over the following 24 hours. If GPT-5.6 Sol is missing from the ChatGPT model picker, that does not necessarily mean the release was reversed: the account may still be waiting for rollout, the plan may not include Sol, or a managed-workspace administrator may need to enable it.

ChatGPT model picker and plan access

GPT-5.5 Instant remains the fast default for ordinary ChatGPT conversations. On eligible plans, GPT-5.6 Sol powers the reasoning-oriented choices rather than replacing every response automatically. OpenAI's current Help Center maps Medium and High reasoning to Sol, while Extra High and Sol Pro availability varies by plan.

Free, Go, and logged-out users do not get GPT-5.6 Sol in standard ChatGPT conversations. Their access differs in Codex and Work: OpenAI lists Terra for Free and Go in Codex, while paid users can choose among Sol, Terra, and Luna in supported product surfaces.

Managed Business and Enterprise workspaces add another layer. Even when a plan supports GPT-5.6, administrators may control model availability. Teams should therefore check the exact workspace and product rather than assuming that access in one personal account proves organization-wide availability.

Sol, Terra, and Luna serve different workloads

OpenAI positions Sol as the flagship model for complex coding, research, science, cybersecurity, computer use, and design. Sol Pro is the highest-capability option for difficult or longer-running work where quality matters more than speed.

Terra is the balanced GPT-5.6 tier for everyday work, and Luna is the fastest and lowest-cost member of the family. Neither Terra nor Luna is a standard selectable model in ordinary ChatGPT conversations. They matter more in Work, Codex, and API routing, where teams can choose a tier that fits latency, cost, and verification needs.

The safest choice is task-based evaluation. A model that performs well on a difficult coding benchmark may still be the wrong default for a high-volume extraction job, and the cheapest tier may create more review work on tasks with subtle correctness requirements.

How to verify GPT-5.6 access

Use the product surface that matters to your workflow:

• In ChatGPT, open the model picker and inspect the available reasoning levels. If Sol is missing, confirm plan eligibility, rollout status, and workspace administration settings.
• In Codex, update the desktop app or CLI before concluding that the account lacks access. Then check which of Sol, Terra, and Luna the workspace exposes.
• In the API, inspect the current model catalog and organization limits. Do not infer API access from the ChatGPT picker.
• In managed workspaces, confirm that administrators have enabled the model family and that data controls match the intended workload.

Migration checks before changing defaults

General availability is a reason to test, not a reason to replace a production model without evidence. Run the same representative tasks against the current default and the GPT-5.6 tier you are considering. Measure answer quality, tool use, structured-output reliability, latency, token consumption, refusal behavior, and total review time.

Keep fallback routing active during the rollout. Coding agents and other long-running workflows should also be tested for compaction, tool-call recovery, destructive-action boundaries, and behavior after interrupted runs. For a reusable evaluation structure, LinkLoot's AI workflow automation guide can help map prompts, tools, approvals, and verification steps.

Sources and update history

OpenAI's July 9 launch page confirms general availability across ChatGPT, Codex, and the API. The GPT-5.6 Help Center article documents current model-picker behavior, plan access, product differences, and gradual rollout. OpenAI's model release notes separately confirm the July 9 ChatGPT rollout.

This article originally tracked the pre-launch signals on July 8. It was updated after OpenAI published the general-availability announcement, so the old limited-preview and launch-watch guidance no longer applies.