Release Watch

GPT-6 Luna Release Date: Rollout Status for API Teams

Jessie

COO

September 20, 2026

6 min read

A GPT-6 Luna release date has not been verified in the official sources checked on September 20, 2026. We found no GPT-6 Luna entry in OpenAI's model catalog or pricing page. Those pages do document GPT-5.6 Luna. A reference to “Luna” without a generation number can therefore point to the existing model.

If you process a queue of documents or classifications, a rumored launch date is not enough to schedule a migration. You need to know which channel has opened, which model it serves and whether it can handle your workload. This page tracks those rollout milestones. For model-ID, pricing and EvoLink access status, use the GPT-6 Luna API page. For the evaluation you can prepare now, use GPT-6 Luna vs GPT-5.6 Luna.

What is confirmed about the Luna rollout?

Milestone	Status checked September 20, 2026	Evidence required to update it
Official launch date	Not verified in the sources checked	A dated OpenAI announcement explicitly naming GPT-6 Luna
Official API reference	No entry found for this generation of Luna	A version-specific model reference and request identity
Official rate	No GPT-6 Luna price row found	A published model-specific rate and billing conditions
Application or account rollout	Scope not independently verified	An official notice naming the product, model and eligible accounts
EvoLink availability	No verified route, rate or activation date available for this tracker	An authenticated request, identity check and usage/billing verification

“Not found in these sources” is a limited observation. It does not prove that no internal trial exists. Equally, a possible trial would not establish general availability, queue capacity or an API price. We will update the rows independently when their evidence changes.

Distinguish GPT-6 Luna from the existing GPT-5.6 Luna

The GPT-5.6 Luna reference identifies gpt-5.6-luna as a model intended for cost-sensitive, high-volume work. That positioning belongs to the documented 5.6 model. It is a useful description of the audience watching for a successor, but it does not prove that GPT-6 Luna will be cheaper or faster.

Check the full identifier before interpreting a search result, dashboard label or release summary. An article that compares GPT-6 Astra with GPT-5.6 Luna contains the words “GPT-6” and “Luna” without discussing a GPT-6 Luna release. The version matters just as much as the tier name.

Do not replace an existing queue's model field with a guessed slug. A webpage address can name a topic before any provider has assigned a callable model ID. Keep your current route explicit and record the candidate separately after the relevant provider documents it.

Which release signals matter to a batch-processing team?

GPT-6 Luna evidence workflow checking source documents and isolating unverified claims before a route is approved

This is an illustration of the verification process, not evidence that the candidate is available.

For repeated workloads, access is only the start of the operational check. Evaluate each new signal for what it actually establishes:

An announcement can establish a name and date. It may say nothing about your account or provider.
A model reference can establish upstream capabilities and supported endpoints. Your gateway still needs its own implementation check.
A successful request can establish access for a particular route and account at that moment. It does not prove sustained throughput.
A priced, observed trial can establish whether your own queue meets its accuracy, latency and budget targets under the tested conditions.

A low token price cannot answer whether retries will consume the saving. A fast single response cannot answer whether the queue meets its deadline at normal concurrency. Treat launch news as the trigger to investigate these questions, not their answer.

Prepare the queue before planning the migration date

Start with a small representative sample from work you already understand. Include ordinary records, missing information, long inputs, ambiguous labels and records your current workflow escalates. Save the expected fields or decisions and the rules for allowing an empty value or an explicit refusal.

Then record three views of the GPT-5.6 Luna baseline:

Correctness: accepted records divided by all attempted records, with schema errors separated from wrong values.
Completion: accepted records per minute and end-to-end P95 time, including queue wait and retry delay.
Spend: the full pipeline bill per accepted record, including failed attempts and any escalation to another model.

These measurements help you decide whether a newly documented route deserves a trial. They also prevent a launch-day demo from becoming a migration deadline. The Luna upgrade guide adds a field-level test matrix and rollback criteria. For present-day choices, see the GPT-5.6 product page and GPT collection.

What will be updated when documentation appears?

This tracker will record a release date only with a supporting source. It will keep application rollout, public API access and EvoLink activation separate, including their respective check dates. If only one channel opens, the other rows will retain their current status.

Model-specific prices and request details will be maintained on the product page. Evaluation results will be added to the comparison guide only after a documented trial, with its sample, settings and limitations. Until then, you can subscribe to the GPT-6 Luna availability alert while keeping your existing queue operational.

FAQ

When will GPT-6 Luna be released?

No date was verified in the official catalog and pricing sources checked on September 20, 2026. We will update this tracker when a model-specific announcement establishes one.

Is the documented Luna model already GPT-6?

The reference checked is GPT-5.6 Luna. A reference to Luna without a generation number should not be interpreted as a new release.

Will GPT-6 Luna launch with another GPT-6 model?

We have no verified schedule establishing a joint launch. Each model and access channel needs its own evidence, even when teams prepare for them together.

Does a low-price claim confirm a release?

No. A price prediction, an old-generation rate or an unrelated provider quote cannot establish a new model's availability.

Can I set a queue migration date now?

You can prepare an evaluation window, but keep the production migration conditional on verified access and measured accuracy, throughput and cost for your queue.