Came across this from OpenAI: GPT-6.1 Terra is now available—a mid-tier model geared toward everyday office work, focused on using a computer by looking at the screen (computer use) and visual reasoning. Officially, it’s a clear step up from GPT-5.6 Terra, while standard API prices have been cut in half.
Key specs: Standard API pricing is about $1 per million input tokens and $5 per million output tokens; on an internal benchmark with error-prone prompts, the factual error rate at low reasoning effort dropped from about 53.6% to about 16.7%; Terminal-Bench 4.0 reaches up to about 50.3% with the highest reasoning effort (versus about 21.5% for the previous generation); and it completes tasks more than three times faster than GPT-6 Astra / GPT-6.1 Sol. It’s now available in ChatGPT Work and Codex for Plus/Pro/Enterprise/Edu users; developers can use gpt-6.1-terra. For complex coding, they still recommend Sol / Astra.
The mid-tier model makes the “look at the screen, click buttons, write things” workflow even more affordable—worth noting to see whether everyday agent tasks become a better deal.
#AI #OpenAI #GPT