OpenAI launches GPT-4o (S2S) model to a small number of users, which seems to be a free version of the real-time speech function

OpenAI recently announced that the advanced real-time voice mode for ChatGPT, which was showcased at their spring conference, will be postponed until July. Initially, this feature will be available for testing by invitation only to ChatGPT Plus subscribers, with plans to extend it to all subscription holders by the fall.

In the meantime, a few users have noticed an unknown new model in the ChatGPT Alpha Models list, named GPT-4o (S2S). This model, for the time being, appears to be similar in use to the existing GPT-4o.

While there has been discussion within the OpenAI community, no official explanation has been provided, leaving the specific purpose of this new model unclear.

Two main speculations have arisen from the discussions:

The first is that S2S might be a rolling release model of GPT-4o, meaning OpenAI continuously updates the model and delivers it to a select group of users for testing.

The second speculation is that S2S might be a free version of the GPT-4o advanced real-time voice mode, allowing free users access to real-time voice features in the future.

The likelihood seems higher for S2S to be a free version of the advanced real-time voice mode, similar to GPT-4o, which is accessible to both paid and free users. However, free users face lower quotas and are temporarily frozen after several interactions, requiring a wait time before regaining access.

Hence, it's plausible that the advanced real-time voice mode will adopt a similar approach, offering more features and higher quotas to ChatGPT Plus subscribers, while providing a limited quota to free users using S2S, essentially serving as a trial version.

Given that it is already July, OpenAI is likely preparing to launch the advanced real-time voice mode for ChatGPT Plus users. If S2S is indeed the free version, ChatGPT free users can also expect invitations to test this feature soon.