OpenAI’s latest AI model targets routine computer work with standard API token prices cut in half compared with its predecessor.
OpenAI launched GPT-6.1 Terra on Monday, Oct. 26, positioning it as a more affordable option for developers and businesses automating routine digital work. The model succeeds GPT-5.6 Terra and focuses on visual reasoning, desktop interaction, coding and workplace communication.
According to the ChatGPT maker, standard API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol. Developers pay $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens.
OpenAI says the model can use connected tools and software to enter invoice details into accounting systems, complete forms and edit videos, alongside interpreting technical drawings. OpenAI also says it performs some professional tasks at a level comparable to its more expensive GPT-6 Astra model.
What OpenAI’s benchmarks show
OpenAI’s benchmark results suggest the update delivers improvements beyond pricing. On Agents’ Last Exam, which evaluates AI performance on complex professional tasks, GPT-6.1 Terra scored 15 percentage points higher than its predecessor at medium reasoning effort while costing about one-third as much.
In its Agents’ Last Exam comparison, OpenAI reports that Terra completes tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol. At maximum reasoning effort, the company also reports roughly nine times lower task costs than Anthropic’s Claude Opus 5.5.
The company also reports fewer factual errors in its internal testing. On OpenAI’s internal benchmark of prompts designed to expose model weaknesses, the error rate at low reasoning effort dropped from 53.6% for GPT-5.6 Terra to 16.7% for the new model. At maximum effort, it fell by 18 percentage points to 4.3%.
Coding performance also more than doubled on Terminal-Bench 4.0, rising from 21.5% to 50.3% at maximum reasoning effort. OpenAI nevertheless recommends GPT-6.1 Sol and GPT-6 Astra for its most demanding coding tasks.
What lower prices mean for AI adoption
The bigger opportunity is making computer-using AI agents economical enough for repetitive, high-volume work. A business processing thousands of invoices or handling routine data entry could benefit from lower per-task costs, particularly when those workflows require multiple steps across software applications.
However, API prices alone do not determine total operating costs. Businesses must also account for repeated model calls, integration work, human reviews and mistakes that require correction. Even a lower error rate does not eliminate the need to verify sensitive financial or operational actions.
For developers, Terra could be a practical middle ground between cheaper, less capable models and premium systems. Choosing it for routine tasks while reserving more expensive models for difficult work could help control AI spending without sacrificing as much capability.
Availability and user impact
GPT-6.1 Terra is available through the OpenAI API under the model identifier gpt-6.1-terra. OpenAI also offers it to Plus, Pro, Enterprise and Edu users through ChatGPT Work and Codex.
For subscribers, OpenAI says the same usage allowance covers more tasks with Terra than with GPT-5.6 Terra, and substantially more than with GPT-6.1 Sol or GPT-6 Astra.
The model retains safeguards developed for GPT-6 Astra, including protections against malicious prompt injections, according to OpenAI. The company also says Terra is better at acknowledging limitations and staying within the scope of a user’s instructions.
What IT teams should test before expanding automation
For businesses, the useful measure is cost per successfully completed workflow. A lower token price can reduce spending, but those savings may shrink if an agent needs repeated attempts or employees spend substantial time correcting its work.
Start with a defined task, such as extracting invoice fields into a review queue or completing forms with supplied information. Test Terra against the model or manual process already in use, using the same inputs and acceptance criteria. Record completion rates, processing time, model spending and the human effort needed to check or repair results.
Evaluate access controls alongside performance. An agent entering accounting data should receive only the permissions required for that workflow, with approval before consequential changes and logs that make its actions traceable. Prompt-injection safeguards should complement those controls.
Expand deployment when the trial shows acceptable accuracy and measurable savings after review costs. For ChatGPT Work and Codex subscribers, the benefit is more tasks within the existing usage allowance; teams should still assess whether the completed work meets their standards.
Want to learn AI tips, tricks, and prompting techniques? TechRepublic readers get free 7-day access to The Neuron Academy, our practical learning platform designed to help professionals use AI more confidently at work. Browse all lessons →
Let us teach you How to Talk to AI for free! Try our six-minute course at The Neuron Academy and learn a few simple ways to write better prompts and get more useful results from AI, or browse our other AI course for free for seven days. Check out all the lessons here →