OpenAI launches GPT-6.1 Terra, a cheaper agent model for computer use
The mid-tier model is built to read screens, operate apps and fill in forms, which OpenAI says it does three times faster than its top models. Its system card also flags rule-breaking in simulations.
Symbolic image · Photo: Coolcaesar / Wikimedia Commons, CC BY 4.0 (cropped and resized)
OpenAI on Monday released GPT-6.1 Terra, the successor to GPT-5.6 Terra, describing it as a "fast and efficient model" that "combines detailed visual reasoning and precise computer use." It targets tasks such as navigating websites, filling in forms and reading technical drawings. The model is available now in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and costs $1 per million input tokens and $5 per million output tokens via the API, half the price of GPT-5.6 Terra and GPT-6.1 Sol.
OpenAI said Terra completes tasks on average more than three times faster than GPT-6.1 Sol and GPT-6 Astra. On the Terminal-Bench 4.0 coding test at maximum reasoning effort it solved 50.3% of tasks, up from 21.5% for its predecessor, while its factual error rate in an internal test at low effort fell to 16.7% from 53.6%. OpenAI still recommends its larger models for the most complex coding work.
The system card also lists risks, SiliconANGLE reported: in nearly 30% of simulated runs the model circumvented restrictions using unauthorized methods, and it took unauthorized actions in 13.75% of cases after receiving messages from other AI agents. Researchers also found signs the model recognized it was being tested.