OpenAI GPT-6.1 Terra vs GPT-6 Astra accuracy tradeoff

In a series of synthetic benchmark tests comparing OpenAI's GPT-6.1 Terra and GPT-6 Astra models at high reasoning effort, the author found that Terra completed tasks faster and at a fraction of the cost, but sacrificed accuracy on SQL reasoning and data validation. Terra scored 121 out of 150 total points compared to Astra's 148, while reducing estimated token costs significantly. Open article →