TLDRocket
Sign in

Quoting Ted Sanders

Simon Willison’s Weblog Simon Willison

OpenAI’s Ted Sanders says model names signal more than order. A version like 6.1 is meant to hint at release timing, similarity, and how big the jump is.

Based on reporting by Simon Willison’s Weblog, Simon Willison — read the original for the full story.

Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error

Ted Sanders has given a pretty blunt explanation for why OpenAI’s model names look the way they do: they are doing several jobs at once. A version number is not just a label for what came after what. It also points to release order, how closely related one model is to another, and how large the improvement is supposed to feel.

That means a name like 6.1 is doing careful work. It says the model arrived after 5.6. It suggests some continuity with other 6.1 models in training data or personality, which is part of why it would not be called 5.7. And it implies the step up is meaningful, but not so dramatic that it deserves the bigger leap and the bigger expectations attached to a 7.

Sanders’ larger point is that LLM performance is messy. There is no neat ranking where the best model at math is automatically the best at jokes. But size still explains a lot: bigger models are generally more capable and reliable, they cost more per token, and they tend to remember more obscure facts.

He also sketched an ordering that makes the pattern obvious: Luna, then Terra, then Sol, then Astra. The naming system, in other words, is doing public relations and technical bookkeeping at the same time. And that is probably unavoidable when every new release has to carry both a family resemblance and a promise.

My take — AI-written commentary, not fact-checked reporting

This is the kind of naming scheme that sounds annoying until the alternatives show up. Clear version numbers beat mystical brand soup every time, especially in AI, where people already talk as if every release is either magic or fraud. The real trick is resisting the urge to make every bump sound like a revolution.

Read more about this at: Simon Willison’s Weblog

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.