1300 166 138 Tell us what you need
News / OpenAI

OpenAI’s GPT 6.1 Terra halves the price of everyday AI work

By Monika Tantau • Published 3 weeks ago
A hand holds a phone showing the OpenAI website next to a vase of pink flowers, as OpenAI launches its cheaper GPT 6.1 Terra model

Key takeaways

• OpenAI released GPT 6.1 Terra on 26 October for ChatGPT Plus, Pro, Enterprise and Edu users and the API
• API prices are $1.42 per million input tokens and $7.11 per million output, half those of GPT 6.1 Sol
• OpenAI says it matches GPT 6 Astra on a computer use test at a tenth of the price per token
• Its error rate on OpenAI's hallucination test fell from 53.6% to 16.7% at low reasoning effort

Most Read

OpenAI has released GPT 6.1 Terra, a faster and cheaper model built for the computer based work that fills a normal day: reading documents, filling in forms, drafting proposals and working through apps on screen. OpenAI says its standard API price is 50% lower than both GPT 5.6 Terra and GPT 6.1 Sol.

At $1.42 per million input tokens and $7.11 per million output tokens, Terra also costs half as much as Claude Sonnet 5.5. It is available now in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and to developers through the API, according to OpenAI’s announcement.

What GPT 6.1 Terra is built for

Terra combines visual reasoning, computer use and clearer writing. OpenAI’s examples include interpreting technical drawings, matching product photos to catalogue entries, editing videos and filling out forms. It reads text and images, writes text, and handles just over a million tokens of context.

The use case OpenAI leans on most is high volume admin, such as reading incoming invoices and entering their details into accounting software. That is the kind of job Australian finance teams already hand to finance automation, and a lower price per document changes the maths on it.

What OpenAI’s own tests show

On Agents’ Last Exam, a benchmark of professional computer tasks, OpenAI says Terra at medium effort scores 15 points higher than GPT 5.6 Terra at about a third of the cost, and performs on par with GPT 6 Astra at a tenth of its price per token. It also claims tasks finish more than three times faster than on Astra or Sol.

Coding improves too. On Terminal Bench 4.0, Terra solves 50.3% of tasks at maximum effort, more than double the 21.5% of GPT 5.6 Terra. OpenAI is clear that Sol and Astra remain the better choice for the hardest coding work, and every figure here comes from OpenAI rather than an independent test.

A phone open on a new ChatGPT chat with example prompts, the app where GPT 6.1 Terra is now available to paying users

Fewer made up answers

OpenAI says it has worked hard on factual errors this year. On its internal set of prompts designed to trip models up, the error rate at low reasoning effort fell from 53.6% with GPT 5.6 Terra to 16.7% with GPT 6.1 Terra, and to 4.3% at maximum effort, close to Astra and Sol.

On safety, OpenAI rates Terra at its Critical threshold for cybersecurity under the Preparedness Framework and says it keeps the protections built for GPT 6 Astra. It also reports stronger resistance to prompt injection, the hidden instructions that try to pull an agent away from its task.

Why Australian businesses should care

For local teams the appeal is volume. Invoice entry, catalogue matching and form filling are repetitive jobs where a cheaper model that can see the screen saves real money. OpenAI bills in US dollars, so what an Australian team pays moves with the exchange rate.

Two details in OpenAI’s developer documentation matter for budgets. Regional data processing adds 10% to the price, which counts for anyone with data residency rules, and prompts longer than 272,000 tokens cost twice as much on input. Teams choosing a tier can compare the options in our guide to GPT 6 Astra, Sol and Luna.

Where Terra fits in the line up

Terra sits below the flagship models on price, not far below them on everyday tasks. Astra stays at the top, and GPT 6.1 Sol remains the stronger all rounder for complex work at twice the cost.

For many businesses the sensible pattern is now a mix: a cheap, fast model like Terra for the bulk of routine work, and a frontier model kept for the hard cases. The next few weeks of independent testing will show how close Terra really gets. Our comparison with Claude Sonnet 5.5 now has those independent scores at every effort.

Scroll to Top