OpenAI's New Terra Model Wants to Drive Your Computer, and It's Half Off
OpenAI just rolled out GPT-6.1 Terra, and the pitch is pretty simple: a cheaper, faster model that's good at actually using a computer. Think filling out forms, editing videos, reading technical drawings and grinding through the boring, repetitive desktop chores nobody wants to do.
The price is the headline. Through the API, Terra costs $1 per million input tokens and $5 per million output tokens. That's half what OpenAI charges for the older GPT-5.6 Terra and for its bigger GPT-6.1 Sol model. OpenAI even claims Terra performs about as well as its GPT-6 Astra model at a tenth of the price per token, and finishes tasks more than three times faster than both Astra and Sol.
The company's own benchmark numbers are big jumps, if they hold up. On a test called Agents' Last Exam, Terra scores 15 points higher than the old Terra at medium effort while costing about a third as much. The error rate at low effort dropped from 53.6% to 16.7%. On the Terminal-Bench 4.0 coding test at max effort, it hit 50.3%, more than double the previous version's 21.5%. Keep in mind these are OpenAI's internal figures, not independent tests, and OpenAI still points people to Sol or Astra for the hardest coding jobs.
It's available now for Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and developers can call it as gpt-6.1-terra.
The bigger story is the price war. Last week Moonshot's Kimi undercut the top U.S. models, and now OpenAI is slashing its own prices. Cheap AI agents that click around your screen for you are clearly where this race is headed.