OpenAI has launched GPT-6.1 Terra, a lower-cost model designed for computer use, visual reasoning, coding, and everyday tasks. The company says Terra can handle activities such as filling out forms, editing videos, interpreting technical drawings, and automating repetitive desktop work while operating faster and costing less than larger models.
“We are introducing GPT-6.1 Terra, a fast and efficient model that combines detailed visual reasoning and precise computational use with clearer, more natural writing,” OpenAI explained in a blog post.
OpenAI described the model as “a huge step up from GPT-5.6 Terra,” saying it excels at weekday communications and computer-based tasks, including writing proposals, interpreting technical drawings, filling out forms, and editing videos.
GPT-6.1 Terra costs less and runs faster
OpenAI says GPT-6.1 Terra costs 50% less through its API than GPT-5.6 Terra and GPT-6.1 Sol. Pricing is $1 per million input tokens and $5 per million output tokens.
In the agent’s final exam, Terra scored 15 percent higher than GPT-5.6 Terra at moderate inference effort while costing about one-third as much.
OpenAI also said Terra costs one-tenth as much per token as GPT-6 Astra while delivering comparable performance. The company said Terra completes tasks more than three times faster than both GPT-6 Astra and GPT-6.1 Sol.
Lower error rates and stronger coding performance
OpenAI’s internal benchmarks indicate that GPT-6.1 Terra also produces fewer errors. At low inference effort, its error rate fell from 53.6% for GPT-5.6 Terra to 16.7%.
“With maximum effort, the error rate decreased by 18 percentage points to 4.3%, roughly matching the error rate of GPT-6 Astra and GPT-6.1 Sol,” OpenAI said.
Coding performance improved as well. Terra scored 50.3% on Terminal-Bench 4.0 at maximum inference effort, more than double GPT-5.6 Terra’s 21.5% score.
OpenAI recommends GPT-6.1 Sol or GPT-6 Astra for the most complex coding tasks.
Availability
GPT-6.1 Terra is now available to ChatGPT Work and Codex users, as well as Plus, Pro, Enterprise, and Edu customers through the API. The model identifier is gpt-6.1-terra.
The overall prevention score can hide what happens after the initial access. If an attacker uses valid credentials, your defenses drop sharply.
The Blue Report 2026 measures defense techniques by technology across 338 million simulations run in customer production environments.
Source: www.bleepingcomputer.com