OpenAI says picking the priciest AI model is not always the smart move

OpenAI says the biggest, priciest AI model is not always the smart pick. Ted Sanders at OpenAI shared a simple rule for choosing among its GPT models. The new GPT-6.1 Terra is his example of why.

What Sanders said

Sanders says Terra is likely better than the larger Astra at technical communication. It roughly matches Astra at image processing and computer use, and it costs much less. In most other areas, Terra falls short.

He adds that model skill has many sides. A model that is great at math will not automatically be good at jokes. Most gaps still come down to size, but not all. Newer, smaller models can beat older, larger ones thanks to better training data or improved design.

Sanders also says OpenAI's model names are hard to simplify. Size, reasoning effort and multi-agent systems (setups where several AI programs work together) are separate dimensions. A single IQ score would hide those details. His rule of thumb:

• Care only about quality: pick Astra.
• A bit price sensitive: pick Sol.
• Very price sensitive: pick Terra or Luna.

What is not known

The material does not give prices, test scores or the tasks behind these comparisons. It also says little about how Sol and Luna differ. Businesses that pay for AI by usage may find a cheaper model handles many of their tasks and trims the bill. Testing it on your own work is the way to find out.