OpenAI launched GPT-6.1 Terra on 26 October 2026, pitching a model built less for abstract feats of reasoning than for the repetitive work that fills a screen-bound day: processing invoices, completing forms, reading technical drawings and editing video. The company set prices at $1 per million input tokens and $5 per million output tokens—50% below GPT-5.6 Terra and GPT-6.1 Sol.¹

The company’s case is that lower cost does not mean lower capability. Terra combines visual reasoning, natural-language writing and the ability to act across applications; OpenAI says it is available through ChatGPT Work, Codex and its API.² Greg Brockman, OpenAI’s president, called the change a qualitative shift: “computer use with terra has crossed a qualitative threshold where it feels faster at using a computer than what i can do unassisted.”Greg Brockman (gdb)³

Sam Altman framed the release as the opening of “this new era of computer use,” rather than merely another benchmark contest.Sam Altman (sama)⁴ The company says Terra scores 15 percentage points above GPT-5.6 Terra on Agents’ Last Exam at medium reasoning effort, for about a third of the cost, while reserving its more expensive Sol and Astra models for the hardest coding and research work.¹

That price cut, however, lands amid a broader market unease. Ramp’s corporate-spending-based token index put prices at roughly 59 cents per million tokens last Friday, down about 35% since late June. Its reading: an apparent price war between OpenAI and Anthropic is unfolding even as usage rises.⁵

AI boosters see the same trend as the path to mass adoption: cheaper computing should pull more work onto the models. Investors are watching a different metric. Usage can surge while spending growth flattens, leaving the industry’s vast data-center investment dependent on demand that is real—but may be getting cheaper faster than it becomes profitable.⁵