OpenAI unveils GPT-6.1 Terra, a cheaper AI model for everyday computer tasks

OpenAI introduced GPT-6.1 Terra, a mid-tier language model aimed at typical computer-based work such as proposal writing, form completion and video trimming. The model is offered through the API under the identifier gpt-6.1-terra, with pricing set at $1 per million input tokens and $5 per million output tokens, roughly half the cost of GPT-5.6 Terra and GPT-6.1 Sol. In internal testing, Terra beat GPT-5.6 Terra by 15 percentage points on the Agents' Last Exam benchmark while using about one-third of the budget, and it matches GPT-6 Astra’s performance at a tenth of the per-token cost.

It also processes tasks more than three times faster than both GPT-6 Astra and GPT-6.1 Sol, and at maximum effort it is over three times as token-efficient as Claude Opus 5.5 at one-fourth the price. Error rates dropped to 16.7% at low reasoning effort and 4.3% at full effort, and on a command-line coding test it solved 50.3% of tasks versus 21.5% for GPT-5.6 Terra. The model meets OpenAI’s Critical cybersecurity threshold and inherits protections from GPT-6 Astra. It became available on Monday for developers via the API and for Pro, Enterprise and Edu users in ChatGPT Work.