Codex users have spotted the name gpt-6.2-astra in usage analytics, even though the model picker still showed GPT-6 Astra. On November 2, TestingCatalog reported the discovery after examining screenshots shared by users.
As of November 4, OpenAI has not announced GPT-6.2 Astra. A line in a usage report alone cannot tell us whether the company is preparing a new model, running a test, or using a different internal label.
What users found
Codex is OpenAI's coding tool: you can ask it to explore a project, make changes, or help diagnose an error. Usage analytics show which models took part in the work. According to users, a line labeled gpt-6.2-astra appeared among the familiar model names.
TestingCatalog describes several screenshots. In one seven-day report the line accounted for 19.8% of usage, while another showed 1.6%. The publication says the entry appeared on October 31 in the reports it examined.
Those percentages describe individual users' reports. They cannot tell us how many people received access to the possible version or what share of all OpenAI requests it handled.
According to TestingCatalog, another shared snippet showed gpt-6.2-astra in the response's model field. That is another report of the same name, but it does not replace a public description from OpenAI.
What people are discussing on X and Reddit
In a post by Lumina cited by TestingCatalog, the user says GPT-6.2 Astra appears in some Codex usage analytics. Other users connect the discovery to a possible test or imminent launch. Those are interpretations: OpenAI has not explained the new label.
In a Reddit discussion predating the discovery, users shared their expectations for future OpenAI models. They discussed speed, usage limits, and whether more affordable models can handle difficult tasks. Some were happy with the current Astra; others wanted similar quality with faster responses and lower usage. The discussion concerns the lineup as a whole and does not establish GPT-6.2 Astra's capabilities.
A new model or an internal name?
Different names in the interface and the usage report could have several explanations. The company might have routed some requests to an updated model. Alternatively, the name could be an internal label that does not correspond to a separate public product.
Both explanations remain speculation. Screenshots reveal neither the model's design nor how its answers compare with GPT-6 Astra. Even a convincing Codex demo cannot establish an improvement without confirming which model ran it and under what conditions.
For me, the key distinction is between the observation and the conclusion: users are discussing the name, but it is too early to call this a full release. An official announcement and comparable tests would give us a better basis for evaluating it.
Why users expect an update
OpenAI has updated the GPT-6 family frequently in recent months. It released GPT-6 Astra on September 3, followed by GPT-6 Sol and Luna on September 22, GPT-6.1 Sol on September 29, and GPT-6.1 Terra on October 26, according to its release list. These models serve different purposes, so the gaps between them do not set a schedule for Astra. The pace does help explain why a new name drew attention.
Competition has also intensified. In Agent Arena, GPT-6 Astra ranked fourth on November 4 for user-confirmed task completion, behind Claude Opus 5.5, Claude Sonnet 5.5, and Gemini 4 Carbon. That is one measure in one leaderboard; it does not mean Astra loses at every task. Rankings and results can change.
There are rumors about the next Claude models, too. Decrypt reports that the names Fable 5.5 and Mythos 5.5 appeared in a third-party catalog and code on an Anthropic page. Neither finding confirms an official release date. A possible competitor release and leaderboard results provide context, but do not prove OpenAI has scheduled GPT-6.2 Astra for the near future.
What OpenAI confirms
The OpenAI model catalog lists GPT-6 Astra under the ID gpt-6-astra. The page we checked has no separate gpt-6.2-astra entry. That describes the public catalog at the time of our check; it does not rule out internal experiments.
OpenAI has not published specifications, pricing, limits, or a release date for the purported GPT-6.2 Astra. It is also unclear whether that version will appear under the same name in Codex or the API. Applying GPT-6 Astra's prices and capabilities to it would be premature.
What to do if you use Codex
For now, rely on the models available in your interface and the official documentation. A new line in a report does not necessarily mean you can select it for regular work or that future requests will use it.
When choosing a model for a project, compare results on your own tasks: ask each model to diagnose the same error or propose a change, then check the code and the cost. If OpenAI announces a new version and its terms, you can run a meaningful comparison.
You can follow available models in the GPTunneL catalog. The presence of GPT-6 Astra there does not mean the purported GPT-6.2 Astra is available: its status remains unconfirmed.