OpenAI launches GPT-6.1 Terra: fast, low-cost computer use at $1/$5 per million tokens, half the price of GPT-5.6 Terra and GPT-6.1 Sol

Background and pain points

Terra is OpenAI's cost-efficient tier below the flagship GPT-6 Astra and GPT-6.1 Sol. The previous GPT-5.6 Terra was cheap, but for long computer-use chains (reading screens, clicking, filling forms) teams often had to move up to Sol or Astra, and costs jumped. For high-volume work such as invoice entry, form filling and screenshot understanding, per-task cost and speed decide whether it reaches production.

Architecture highlights

OpenAI did not disclose parameter counts or training details, only product specs: text and image input, text output, a 1,050,000-token context (922,000 max input, 128,000 max output), and a Jun 1 2026 knowledge cutoff. reasoning.effort offers none, low, medium (default), high, xhigh and max, and tool calling requires the Responses API. OpenAI says it inherits GPT-6 series alignment work, stays in scope more often, discloses blockers, and resists prompt injection well. It also says the model meets the Critical cybersecurity threshold under its Preparedness Framework and keeps the GPT-6 Astra safeguards.

Benchmarks

All figures are OpenAI's own and not yet independently reproduced. On Agents' Last Exam it scores 15 points above GPT-5.6 Terra at medium effort for about one-third the cost, and matches GPT-6 Astra at one-tenth the price per token. At max effort it is over 3x as token-efficient as Claude Opus 5.5 at a quarter of the price per token, for about 9x lower cost per task, and it finishes tasks over 3x faster than Astra and Sol. It solves 50.3% of Terminal-Bench 4.0 at max effort versus GPT-5.6 Terra's published 21.5%, though OpenAI still recommends Sol or Astra for the hardest coding work. On an internal error-inducing prompt set, the low-effort error rate fell from 53.6% to 16.7%, and max effort reached 4.3%. On MMMU-Pro it beats GPT-5.6 Terra at every effort level and nearly matches Astra and Sol.

Getting started for developers

Switch the model name to gpt-6.1-terra. Pricing is $1 input, $0.05 cached input and $5 output per million tokens; requests above 272K input tokens are billed at 2x input and 1.5x output for the whole request, so budget long-context jobs carefully, and use Batch/Flex at half price for offline work. It suits high-volume screenshot understanding, form and back-office automation and drafting, while complex refactors still belong on Sol or Astra. Plus, Pro, Enterprise and Edu users can pick it in ChatGPT Work and Codex today, and OpenAI says the same allowance covers more tasks.