OpenAI Releases GPT-6.1 Terra, Priced 50% Lower Than GPT-5.6 Terra
OpenAI has upgraded its mid-sized model Terra from GPT-5.6 to GPT-6.1, focusing on speed, computer operation, and more natural writing, at a price 50% lower than GPT-5.6 Terra. On OpenAI's internal error-prone prompt benchmark, the error rate at low reasoning effort dropped from 53.6% to 16.7%, and at the highest reasoning effort the error rate fell by 18 percentage points to 4.3%. The model is available starting today through ChatGPT Work, Codex, and the API.
Why read this
AI score 75
The original text provides the price reduction and error rate changes of the new model relative to its predecessor, allowing readers to assess the cost-effectiveness trend of mid-sized models.
Topics
Body
Earth is so back. Last month when OpenAI released GPT-6 models of Sol and Luna, its medium model Terra was excluded from the update. Now OpenAI has upgraded Terra from GPT-5.6 to GPT-6.1, touting speed and computer use at a lower price.
GPT-6.1 Terra is OpenAI’s new medium-sized model that balances intelligence and affordability
OpenAI currently has four main model classes, ranging from lowest price and capability to highest: Luna, Terra, Sol, and Astra.
Today’s GPT-6.1 Terra model release sounds like a great solution for doing more work for longer without costing as much.
Here are the high-level details from the announcement:
We’re introducing GPT-6.1 Terra, a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing. A significant step up from GPT‑5.6 Terra, it excels at the communication and computer-based tasks that fill the workday, from drafting proposals and interpreting technical drawings to filling out forms and editing videos.
OpenAI highlights that GPT-6.1 Terra pricing is actually 50% lower than the inferior version that it replaces, GPT-5.6 Terra, which makes it ideal for automating computer work tasks for less.
In terms of intelligence and accuracy, OpenAI says Terra is a step forward in these areas as well:
On our internal benchmark of error-inducing prompts, the error rate at low reasoning effort fell from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra, a drop of nearly 37 percentage points. At maximum effort, the error rate fell by 18 percentage points to 4.3%, nearly matching the rates for GPT-6 Astra and GPT-6.1 Sol.
The new GPT-6.1 Terra model is available today through ChatGPT Work, Codex, and the API. You can learn much more about the new medium-sized model here.
Do more with your Apple products
AirPods 5 | Apple’s newest wireless headphones
Apple AirTag 2 | Add Find My tracking to keys, bags, bikes, more
AirPods Pro 3 | Apple’s best wireless headphones
Wireless CarPlay Adapter | No more plugging in required