OpenAI skips naming its reasoning model 'o2' to 'o3' for the silliest reason

OpenAI is expected to release yet another reasoning-excelling model after the success of the o1. But, it won’t be called the “o2” and instead, it’ll be called the “o3” model.

The Information has exclusive information that the Microsoft-backed company fears a potential lawsuit from a British network company, O2. That’s a potential trademark conflict.

Sources are also saying that the AI startup is using Orion, its latest pre-trained model, to develop o3 and has found initial success in fields like coding, math, and science, with a pilot charging $200 per month.

And if the clue is not clear enough, Sam Altman, OpenAI’s boss, just posted two cryptic tweets.  

“Ho ho ho, see you tomorrow”, he says, to which he replies again to himself, “Should have said oh oh oh.”

Compared to other models, OpenAI o1 gives more thoughtful responses and reasoning through problems like humans. It comes with a “mini” version too that’s more cost-effective and faster for coding tasks.

You can try the model out on ChatGPT already even as a free user using the model picker. As for developers, the company also launched o1’s API, which is priced at $1,000 per month only for Tier 5 users. It includes function calls, structured outputs, real-time updates, and a preference fine-tuning method.

OpenAI has released a lot of exciting features during the ongoing 12-day “shipmas.” Sora, its video-generating model, has finally launched after months of anticipation. But, not too long after, Google soon rose from being an AI laughingstock (the Gemini fiasco, remember?) to challenging the hotly-anticipated model with Veo 2.