OpenAI Released Affordable GPT-6.1 Terra Model for Routine Tasks

OpenAI has introduced the new GPT-6.1 Terra language model, designed for routine tasks at a reduced cost. The model is intended for drafting, analyzing technical drawings, filling out forms, and basic video editing.

The cost of using the model via API is significantly lower than previous versions. Priced at $1 per million input tokens and $5 per million output tokens, it is twice as cheap as GPT-5.6 Terra and GPT-6.1 Sol.

The new model's performance demonstrates high results compared to flagship solutions.

• In the Agents' Last Exam benchmark at the medium reasoning level, the model shows results comparable to GPT-6 Astra, while its token cost is 10 times lower.

Task execution speed exceeds that of Astra and Sol models by more than three times. In the Terminal-Bench 4.0 test, the model successfully handles 50.3% of tasks, compared to 21.5% for its predecessor.

At the maximum reasoning level, using the model costs approximately 9 times less than applying Claude Opus 5.5 per task.

Answer quality has improved significantly: the error rate on tricky questions at the low reasoning level dropped from 53.6% to 16.7%.

Despite its high efficiency, for solving the most complex programming tasks, the company recommends using Sol and Astra models.

Model availability is provided for Plus, Pro, Enterprise, and Edu subscribers via ChatGPT Work and Codex services, as well as through the open API.

The model has reached a critical threshold in cybersecurity according to OpenAI's internal risk scale. To ensure safety, it is protected by the same means as the flagship Astra model.