They are saying that naming is hard, but can you be any worse at naming than OpenAI is?
One the one hand, Sora is a text-to-video model. On the other hand, Sora is an web UI that allows both using a text-to-image and a text-to-video model.
One the one hand Codex is a model created years ago and on the other hand it’s a web UI with a model released in May 2025.
I have no idea what the o in o1/o3/4 stands for.
Confusingly I believe o stands for “Omni” in the context of GPT-4o, since it’s “omni-modal”. Based on some quick googling, the o in o1/o3/o4 seems to emphasize that o1 was resetting the counter back to 1 (so it’s more like zero1).
Given that o1-pro and o3-pro are models that can’t deal with attachments and analyze images, while GPT 4.5 (that has no o in the name) can, that seems to make the naming convention even more awkward.
They are saying that naming is hard, but can you be any worse at naming than OpenAI is?
One the one hand, Sora is a text-to-video model. On the other hand, Sora is an web UI that allows both using a text-to-image and a text-to-video model.
One the one hand Codex is a model created years ago and on the other hand it’s a web UI with a model released in May 2025.
I have no idea what the o in o1/o3/4 stands for.
I’ve heard it stands for “Omni”
Confusingly I believe o stands for “Omni” in the context of GPT-4o, since it’s “omni-modal”. Based on some quick googling, the o in o1/o3/o4 seems to emphasize that o1 was resetting the counter back to 1 (so it’s more like zero1).
Given that o1-pro and o3-pro are models that can’t deal with attachments and analyze images, while GPT 4.5 (that has no o in the name) can, that seems to make the naming convention even more awkward.