OpenAI has also uncovered a mysterious model.
On October 18, developers discovered that three unfamiliar names suddenly appeared in OpenAI's API model list:
The aurora-red-alpha, aurora-green-alpha, aurora-blue-alpha. All three models are named after the aurora, which correspond to the three colors of red, green, and blue.
This is the first time this has happened.
However, OpenAI's withdrawal is a quick one, and according to netizens, the relevant items will soon disappear from the model list.
Some developers tried to call these three models through third-party model list tracking tools, and with no exceptions, they all returned HTTP 503 errors and pointed out:
No available capacity was found for the model.
This indicates that the identifiers of the three models may have already appeared in the relevant system, but cannot be used properly at this time.
This is the first time this has happened.
It is noteworthy that this discovery of a list of OpenAI models maintained by a third party is not an official announcement. Therefore, the exact nature of these models, whether they will be officially released, is currently unknown.
Some netizens speculate that these models are likely to be internal checkpoints for later versions of the GPT-6 Astra, specifically for the strengths of Fable 5.5, creative coding, long-range agents, visual / 3D.
There are also netizens who speculate that three models correspond to OpenAI and three image models.
Some netizens even claim to have experienced these mysterious models on Image Arena's image-generating test platform.
In the comments section, the netizens are not looking, and they have started the modeling mode.
Some netizens speculate that the three models appeared too suddenly, disappeared too quickly, the hands were a little slower, even the shots couldn't catch up, the three sets of red, green and blue, withdrawn faster than the book.
Another user laughed and said: "Bad, three models, enough Tibo water for three days".
Some netizens say that every time OpenAI releases a major version, mysterious models always appear in the interface in advance, and similar "accidental leaks" are quickly reserved for programs.
However, it is still unclear what exactly these three Aurora models are.
As of now, OpenAI has not responded officially. Are they early test versions of a new generation of models, or traces left behind by internal experiments?
© THE