OpenAI launched GPT-6.1 Terra on October 26, introducing a model focused on computer use, visual reasoning, coding and everyday work. According to the company, it delivers faster agent performance while cutting API pricing by 50% compared with GPT-5.6 Terra and GPT-6.1 Sol.
The release targets tasks such as completing forms, interpreting technical drawings, drafting proposals and editing videos. Additionally, OpenAI positions the model as a more affordable option for automating repetitive desktop activities.
GPT-6.1 Terra Cuts API Pricing
OpenAI lists API pricing at $1 per million input tokens and $5 per million output tokens. These rates represent half the price of GPT-5.6 Terra and GPT-6.1 Sol, potentially reducing costs for developers running frequent automated tasks.
However, per-token pricing and the total cost of completing a task measure different things. OpenAI separately reports that Terra achieved stronger results than its predecessor on an agent benchmark while costing approximately one-third as much.
On Agents’ Last Exam, the model scored 15 percentage points higher than GPT-5.6 Terra at medium reasoning effort. The company also reports performance comparable to GPT-6 Astra on that evaluation, at one-tenth of Astra’s price per token.
Meanwhile, OpenAI says Terra completed tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol. These figures reflect the company’s reported evaluations rather than a guarantee for every workload.
Internal Tests Show Fewer Errors
OpenAI reports a substantial reduction in errors on its internal benchmark. At low reasoning effort, the error rate fell from 53.6% for GPT-5.6 Terra to 16.7% for the new model.
At maximum reasoning effort, the company recorded an error rate of 4.3%, an improvement of 18 percentage points. According to OpenAI, that result approaches the rates achieved by GPT-6 Astra and GPT-6.1 Sol.
Nevertheless, the figures describe performance on a specific internal evaluation. They do not establish that the model will produce accurate answers across every subject or task.
Coding Results Improve Alongside Computer Use
GPT-6.1 Terra scored 50.3% on Terminal-Bench 4.0 at maximum reasoning effort, compared with 21.5% for its predecessor. That result more than doubles the earlier model’s score and supports OpenAI’s claims of improved coding performance.
Despite those gains, the company continues to recommend GPT-6.1 Sol or GPT-6 Astra for the most complex coding assignments. Terra instead combines its coding improvements with visual interpretation, computer interaction and writing capabilities for routine work.
At launch, OpenAI made the model available to Plus, Pro, Enterprise and Edu users through ChatGPT Work and Codex. Developers can also access it through the API using the identifier gpt-6.1-terra.
0 responses to “GPT-6.1 Terra Launches With Faster Agents, Lower Costs”