OpenAI explains GPT model sizing, naming and selection trade-offs
OpenAI's Ted Sanders says LLM performance is multidimensional, so a model that excels at mathematics may not lead at jokes or other tasks. He adds that model size generally correlates with performance, reliability and recall of esoteric facts, while higher-size models also cost more per token. OpenAI's models are ordered Luna < Terra < Sol < Astra, but Sanders notes that newer, smaller models can outperform older, larger ones through better training data or architectures.
The guidance helps users interpret GPT model names and choose between quality, reliability and token cost, while showing why a single ranking or IQ score cannot describe every capability.
A quote from Ted Sanders
Simon Willison’s Weblog
Sponsored by: Greptile — Your code reviewer uses Computer Use to catch bugs by opening your app and clicking through it. Now you can configure its test environment and define the key user journeys it tests. Try it for free
28th October 2026
In general, LLM performance is incredibly multidimensional. There's no guarantee of a fixed ordering where the model that is best at math is also the model that's best at jokes. Despite that complexity, much of the variation in performance is explained by just a single variable: model size. In general, models that are bigger will (a) be more performant / reliable, (b) cost more per token, and (c) have better recall of esoteric facts. As you note, our models are ordered: Luna < Terra < Sol < Astra [...] These days, version number signifies a few things at once: the order of release, the relationship to other releases, and the magnitude of improvement. So something like 6.1 would indicate it was launched after 5.6, may share some similarities to other 6.1 models in terms of training data / personality (which is why it's not 5.7), and isn't a giant enough leap to merit the headlines and expectations that would come with a 7.