OpenAI Launches GPT-6.1 Terra at Half the Price, Pitching It for Computer Use and Everyday Work
A cheaper, faster model arrives in ChatGPT Work, Codex and the API. OpenAI's own tests show big gains, but the error figures are internal and the company still sends hard coding elsewhere.
Outspoken Digest Technology Desk
Tuesday, October 27, 2026/2 min read
OpenAI has put out a model that is meant to be used rather than admired. GPT-6.1 Terra, announced on Monday, is a smaller and cheaper system built for the unglamorous end of agent work: filling in forms, editing video, reading technical drawings and driving a desktop. It costs half what its sibling models do, and OpenAI says it is much faster.
The company calls it "a fast, efficient model that combines detailed visual reasoning and precise computer use," in a line quoted by BleepingComputer. Every performance number below comes from OpenAI, not from an outside lab.
What is GPT-6.1 Terra and who can use it?
Terra is rolling out to Plus, Pro, Enterprise and Edu users in ChatGPT Work and in Codex, and developers can call it through the API as gpt-6.1-terra. According to Decrypt, free ChatGPT users stay on the lower-tier GPT-5.6 Luna. The Terra name is not new: it appeared in limited preview on 26 June as part of the GPT-5.6 family and opened to the public on 9 July, so this is a new generation of an existing tier.
How much does GPT-6.1 Terra cost?
Standard API pricing is $1 per million input tokens and $5 per million output tokens, which OpenAI says is 50 percent below both GPT-5.6 Terra and GPT-6.1 Sol, BleepingComputer reports. The cut matters because computer-use agents burn through tokens: every screenshot read and every click decided is a billed step, so a halved price changes which routine jobs are worth automating.
How fast and how accurate is it?
OpenAI's headline claims, as relayed by BleepingComputer and Decrypt:
- On Terminal-Bench 4.0 at maximum reasoning effort, Terra solved 50.3 percent of tasks, against 21.5 percent for GPT-5.6 Terra.
- On Agents' Last Exam at medium effort it scored 15 percentage points higher than GPT-5.6 Terra at about a third of the cost, and it matches GPT-6 Astra at roughly a tenth of the per-token price.
- It finishes tasks more than three times faster than both GPT-6 Astra and GPT-6.1 Sol.
- On an internal test built from prompts designed to trick models into mistakes, the error rate at low effort fell from 53.6 percent to 16.7 percent, and at maximum effort it fell by 18 points to 4.3 percent.
Decrypt is blunt about the limits: the error test is OpenAI's own and has not been independently verified. A benchmark designed by the vendor to show a vendor's gains deserves a pinch of salt until third parties repeat it on their own tasks.
Is Terra a replacement for OpenAI's bigger models?
No, and the company says so. OpenAI still recommends GPT-6.1 Sol or GPT-6 Astra for the most complex coding and in-depth research. Terra is pitched as the everyday model, the one that drafts replies, works through office software and handles the repetitive tasks that make up most of a typical agent bill.
That split is the real story. The expensive frontier model is becoming the specialist, while a cheaper, quicker tier does the volume work. If independent testing backs OpenAI's numbers, the next question for rivals is not who has the smartest model, but who can offer something almost as good at a tenth of the price.
Published in The Outspoken Digest
Editorial desk
Outspoken Digest Technology DeskSoftware, hardware, artificial intelligence and what they change for everyone else.
Newsletter
The Digest, in your inbox
One edition, sent when it is ready. No noise, and your address is never passed on.
Read Next
More Technology →
Can Gravitational Fields Break and Reconnect? A New Physical Review D Paper Says Einstein's Equations Allow It
Oct 28, 2026/2 min read
Google DeepMind Unveils Genie 4, a World Model That Now Generates Sound, for Trusted Testers
Oct 28, 2026/2 min read
Meta's $25 Billion Share Sale Drew Five Times the Demand, With Nvidia and AMD as Cornerstone Buyers
Oct 28, 2026/2 min read
Anthropic Files Its IPO Prospectus: $16.3 Billion in Revenue, an $80.7 Billion Loss and a Nasdaq Ticker
Oct 28, 2026/3 min read