OpenAI Astra Leak Points to New Model Testing

An unconfirmed OpenAI Astra leak claims that early preparations for the rumored model have been spotted again, with vega-alpha and ultima-alpha reportedly appearing in testing at the same time. The claim was shared by @M1Astra, which described vega-alpha as a new model identifier.

What the leak claims

The post names two models currently said to be in testing:

• vega-alpha, described in the post as new

• ultima-alpha

The wording suggests that both identifiers were observed simultaneously. However, the post does not establish what either model can do, how they are related to Astra, or whether they represent separate products, internal versions, or test configurations.

Why the report remains unconfirmed

The post is not an official OpenAI announcement. It presents the information as an observation of early preparations, so the reported connection to Astra should be treated as a possibility rather than a confirmed product detail. Model names seen in testing can also change before a public release, making it premature to infer a final name, feature set, launch date, or availability from these identifiers alone.

The attached image shows a stylized OpenAI logo against a dark background. It does not visibly document the model testing described in the post, so the claim rests on the post’s reported model names rather than on independent visual evidence.

For now, the useful takeaway is narrow: the leak points to possible ongoing internal work involving vega-alpha and ultima-alpha, with vega-alpha identified as a new entry. It does not confirm that OpenAI is preparing a public Astra launch or reveal what the rumored model may ultimately become.