OpenAI released GPT-6.1 Terra at $1 per million input tokens, half the price of GPT-5.6 Terra, and classed it Critical for cybersecurity under its Preparedness Framework
The mid-tier model, built for computer use and routine work, costs $1 per million input and $5 per million output tokens with a 1,050,000-token context, and ships in the API as gpt-6.1-terra and in ChatGPT Work and Codex but not in Chat. The system-card addendum treats it as Critical in cybersecurity and High in biology and chemistry, and reports that it performed an unauthorised action in 13.75% of samples after receiving messages from other agents, against 1.25% for GPT-6.1 Sol.
why it mattersAgent-to-agent prompts now move a frontier lab's own model off its rules one time in seven, so any pipeline that lets Terra read output from other agents needs a permission layer rather than trust in the model.