OpenAI launches GPT-6.1 Terra for cheaper, faster everyday computer work

Terra had missed OpenAI's previous round of model upgrades. According to 9to5Mac Apple, Sol and Luna received GPT-6 updates last month, while the medium-sized Terra remained on GPT-5.6.

The new release brings Terra into the GPT-6.1 family. It occupies the middle of the range rather than replacing OpenAI's largest models, with The Register describing it as a budget-priced, general-purpose option below the high-end Sol and Astra offerings.

On October 26, OpenAI released GPT-6.1 Terra as a replacement for GPT-5.6 Terra, according to The Next Web. Its stated focus is routine computer work, rather than simply answering questions in a chat window.

Visual reasoning and computer use were central to the announcement, Simon Willison reported. Those capabilities are intended to help the model interpret what appears on a screen and work with computer interfaces. Digital Trends also identified coding and professional work as priorities.

Price and speed were the other major selling points. BleepingComputer reported claims of three-times-faster AI agents and 50 percent lower pricing. These are launch claims, not evidence in the supplied coverage of equivalent gains in every task.

By October 27, DEV Community reported that Terra had also appeared on OpenRouter, listing a 1.1 million-token context window and prices of $1 for input and $5 for output per million tokens.

Terra is available to subscribers in ChatGPT Work and Codex, and through the API, according to AI on vc.ru. That gives the release both a direct workplace audience and a route into software built by outside developers.

The proposition is a more capable middle-tier model at a lower cost. SiliconANGLE describes it as an upgrade costing around 50 percent less than its predecessor. The supplied reporting, however, offers little independent testing to establish how consistently its advertised improvements translate into completed work.

The practical test is whether lower usage prices and faster agents also mean lower costs for successfully completed tasks. Open questions include how reliably Terra handles unfamiliar interfaces and whether its claimed accuracy gains hold outside demonstrations. The supplied coverage does not settle those questions.