When Was GPT-5.6 Released?
When was GPT-5.6 released? Check the official release date, Sol/Terra/Luna models, API, ChatGPT, and Codex availability.
Dora is coming. I checked this the way I would update a model registry: date first, rollout notes second. If the question is when GPT-5.6 was released, the API answer is July 9, 2026. OpenAI’s official API changelog lists the GPT-5.6 family on Jul 9, 2026, for v1/responses, v1/chat/completions, and v1/batch.
That does not mean every ChatGPT, Codex, or workspace account saw the same thing at the same hour. API release and product rollout are separate records. I would keep them separate in any internal doc.
GPT-5.6 Release Date
Official API changelog date and model family names
OpenAI released the GPT-5.6 model family to the API on July 9, 2026. The changelog names three model IDs:
| Model ID | Registry note |
|---|---|
gpt-5.6-sol | Frontier capability |
gpt-5.6-terra | Balance of intelligence and cost |
gpt-5.6-luna | Efficient, high-volume workloads |
gpt-5.6 | Alias that routes to gpt-5.6-sol |
For release timing, cite the changelog. For model IDs, context, tool support, and pricing fields, cite the OpenAI models page. Those are different jobs. Mixing them creates messy registry notes later.
The short answer stays simple: the GPT-5.6 release date for the OpenAI API is July 9, 2026.
Why ChatGPT rollout and API release can differ
An API release is not the same thing as a ChatGPT model-picker rollout. It is also not the same thing as Codex availability inside the desktop app, CLI, IDE extension, or cloud.
OpenAI’s workspace model availability docs say model access depends on product surface and authentication boundary. That is the sentence I would use to stop vague rollout claims from spreading.
A clean internal note would look like this:
| Surface | Status wording to use |
|---|---|
| API | Released July 9, 2026, per OpenAI API changelog |
| Batch | Listed in the same changelog under v1/batch |
| ChatGPT | Verify in current plan and workspace |
| Codex | Verify in current Codex docs and account |
| Dashboard | Verify inside the project before changing defaults |
What Was Released
Sol, Terra, Luna, and the gpt-5.6 alias
OpenAI’s model guidance frames Sol as the flagship model, Terra as the balanced model, and Luna as the efficient option. The gpt-5.6 alias routes to Sol.
That alias detail matters. If a team wants stable cost or latency behavior, it may prefer pinning gpt-5.6-terra or gpt-5.6-luna instead of using the family alias. If a team wants the flagship default, the alias is easier. Different choice. Different risk.
I would not write “OpenAI GPT 5.6 is one model.” It is a family. Sol, Terra, and Luna are the actual names to track.
API, Batch, ChatGPT, and Codex availability notes
The API changelog confirms Responses API, Chat Completions API, and Batch API coverage. That is enough for the release-date answer.
Codex needs its own note. The current Codex models page lists 5.6 Sol, 5.6 Terra, and 5.6 Luna, including command examples such as codex -m gpt-5.6-sol. That supports a Codex availability note, but it still should not be merged with the API release line.
Do not invent a GPT-5.6 Codex model ID unless OpenAI publishes that exact slug. The docs I checked show GPT-5.6 models used in Codex surfaces. That is not the same as a separate gpt-5.6-codex model.
What Builders Should Verify
Account access, rate limits, regions, and dashboard visibility
Before changing a production default, I would verify five things in the actual OpenAI project:
- the project can call
gpt-5.6-sol - the project can call
gpt-5.6-terra - the project can call
gpt-5.6-luna - Batch jobs work for the intended workload
- dashboard rate limits are high enough for expected traffic
OpenAI’s rate limits guide says limits vary by model and are set at organization and project level. So do not copy another team’s numbers. Check the dashboard.
Model IDs and plan-level eligibility
The model registry should store exact IDs, not display labels. Use gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna. If the app uses gpt-5.6, add a note that the alias routes to Sol as of the current docs.
Plan-level eligibility belongs in another field. API key access, ChatGPT plan access, Codex access, and workspace policy are separate controls. One fewer assumption. That counts.
FAQ
Who should update the internal model registry?
The platform owner or AI infrastructure owner. Product can request the update, but engineering should verify model IDs, endpoints, rate limits, fallback behavior, and dashboard visibility.
What source should teams cite for release timing?
Cite the OpenAI API changelog. For the exact question “when was GPT-5.6 released,” use July 9, 2026 for the API release date.
How should teams handle staggered rollout claims?
Write the claim in two parts. “GPT-5.6 API released July 9, 2026.” Then separately: “ChatGPT, Codex, and workspace availability may depend on plan, product surface, and account settings.”
That is the clean version. When was GPT-5.6 released? For the OpenAI API, July 9, 2026. Everything after that is access verification.
Previous posts:
