Ted Sanders from OpenAI explained the place of GPT-6.1 Terra in the company’s lineup. According to The Decoder’s published description, the smaller model outperforms the larger Astra in technical communication and costs significantly less, but trails it in most other areas.
The key takeaway for model selection is not to judge by size alone. Sanders separates model size, reasoning effort, and multi-agent systems as distinct parameters that cannot be reduced to a single “IQ” metric.
The source is presented as metadata and a short synopsis by The Decoder, rather than the full text of the material or independent confirmation. Therefore, details of comparisons, the range of tasks, and the exact cost differences remain open questions.
editorial commentary
Why it matters
Likely practical consequence — companies will evaluate models by task type and cost, not by a single conditional rating. The next observed signals will be published OpenAI or independent tests with methodology, tasks, and prices. Substantial uncertainty remains due to the absence of the full text and primary data.