OpenAI’s Ted Sanders Offers a Rule of Thumb for Choosing GPT Models

What happened

OpenAI’s Ted Sanders says model size explains many performance differences, but model names also reflect distinct dimensions such as reasoning effort and multi-agent systems. He recommends Astra for quality, Sol for some price sensitivity, and Terra or Luna for high price sensitivity. Sanders says GPT-6.1 Terra likely surpasses the larger Astra in technical communication and roughly matches it in image processing and computer use at a much lower price, while trailing in most other areas.

Read next

• More changes in today’s briefingIntroducing GPT-6.1 Terra

The story so far

OpenAI’s Ted Sanders offered a rule of thumb for choosing GPT models: pick Astra when quality is the sole priority, Sol when price matters somewhat, and Terra or Luna when price matters a lot. Sanders said model size explains many performance differences, but model names also reflect distinct dimensions, including reasoning effort and multi-agent systems. His comparison leaves room for task-specific choices: he said GPT-6.1 Terra likely surpasses the larger Astra in technical communication and roughly matches it in image processing and computer use at a much lower price, while trailing Astra in most other areas.