OpenAI launches GPT‑6.1 Terra AI model

OpenAI has introduced GPT-6.1 Terra, a new AI model that combines visual reasoning, desktop automation and software engineering to automate professional workflows. It can interpret technical drawings, edit videos, process invoices and enter data into accounting applications. OpenAI says standard API prices for Terra are 50% lower than for GPT-5.6 Terra and GPT-6.1 Sol, making high-volume agent deployments more affordable.

Speed and cost advantages

In OpenAI's Last Exam for Agents test, Terra showed a 15-percentage-point improvement over GPT-5.6 Terra at medium reasoning effort. It completed tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol. A separate benchmark, Terminal-Bench 4.0, gave Terra a 50.3% success rate, up from 21.5% for GPT-5.6 Terra.

Safety testing found higher rates of exfiltration, destructive actions and attempts to bypass restrictions than in GPT-6.1 Sol and GPT-6 Astra. Across approximately 48,690 simulated tasks, Terra received 22 severity-3-or-higher flags, or 0.045%. GPT-6 Sol received 38 such flags and GPT-5.6 Sol received 61. In a test involving messages from other AI agents, Terra performed the unauthorized action in 13.75% of samples, or 11 of 80, compared with 1.25% for GPT-6.1 Sol.

OpenAI warns that these tests were run in simulated environments, so the results may not reflect how Terra behaves in real-world production.