OpenAI launches GPT-6.1 Terra with half-price API rates
OpenAI's latest work model targets computer use, visual reasoning and writing, with standard prices of $1 input and $5 output per million tokens.
OpenAI is cutting the standard API price of its Terra work model in half with GPT-6.1 Terra, a new release aimed at computer use, visual reasoning and natural writing. The company’s October 26 announcement lists $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens.
OpenAI says those standard prices are 50% lower than GPT-5.6 Terra and GPT-6.1 Sol. This is a token-price comparison, not a guarantee that every completed task will cost half as much: the number of tokens and tools a workflow uses still matters.
A model for operating everyday software
The launch examples include reading invoices and entering them into accounting software, drafting proposals, and working with forms and video. OpenAI says the model is available through the API as gpt-6.1-terra and is rolling out to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex.
Its pitch is improved capability at a price suitable for repeated work. The company reports gains in agent and terminal benchmarks, but those results are OpenAI’s measurements. AITrending has not independently reproduced them. The announcement itself says Sol and Astra remain better suited to the most complex coding tasks.
Price is clear; task performance still needs checking
The public English Hacker News discussion includes debate about benchmark selection and the difference between advertised token prices and cost per successful task. Those comments establish a live discussion, not a consensus or a replacement benchmark.
For teams choosing a default work model, the concrete new option is a lower standard rate for an available API model. Whether that makes it the best fit depends on the actual software workflow and required reliability. X API coverage was unavailable during this review; public English discussion and original OpenAI sources were checked instead.