OpenAI has updated its solar system with a new member. The developer has unveiled GPT-6.1 Terra, the latest version of its balanced mid-range AI model, succeeding GPT-5.6 Terra, and featuring improvements in speed and efficiency. What the GPT-6.1 Terra introduction says Within OpenAI's model hierarchy, Terra combines competitive performance with a more moderate cost, making it ideal for everyday work. This new version maintains this balance between speed, capacity, and cost, but takes it to a new level. In addition to being faster and more efficient than GPT-5.6, this new model incorporates greater visual reasoning capabilities. Furthermore, GPT-6.1 excels at using the computer to execute complex, multi-step tasks and also at handling AI agents to use code in scientific research. OpenAI notes that, with average reasoning skills, GPT-6.1 Terra was able to “convert architectural plans into a 3D model and review an animated film shot by shot to create a timecode record for a restoration team,” achieving “a score 15 percentage points higher than GPT-5.6 Terra at approximately one-third the cost.” Other tasks in which GPT-6.1 Terra excels are software engineering and writing, producing clearer and more natural-sounding text. Furthermore, if we analyze the Intelligence Index of the independent analysis and comparison platform Artificial Analysis, we observe that this mid-range model already ranks tenth, behind GPT-6.1 Sol. Fewer hallucinations and more guarantees The OpenAI development team has also invested significant effort in reducing hallucinations and factual errors in this AI. “In our internal benchmark test with error-inducing prompts, the error rate with low reasoning effort dropped from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra, a reduction of nearly 37 percentage points. With maximum effort, the error rate fell by 18 percentage points to 4.3%, nearly matching the rates of GPT-6 Astra and GPT-6.1 Sol,” the company stated. Regarding security, GPT-6.1 Terra features the same protections developed for the flagship GPT-6 Astra model. During testing conducted by OpenAI, the model demonstrated resistance to prompt injection attacks (malicious instructions designed to divert it from its task). Furthermore, AI tends toward honesty, as “GPT-6.1 Terra is more likely than GPT-5.6 Terra to stay within the scope of the task and honestly inform the user about its limitations and potential obstacles.” GPT-6.1 Terra Availability The GPT-6.1 Terra model has already begun rolling out to all users on the Plus, Pro, Enterprise, and Edu plans in ChatGPT Work and Codex. Developers can now use this model through the OpenAI API at a standard price of $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens.