OpenAI has introduced GPT-6.1 Terra, a model designed to understand what is on screen, act across applications and write naturally. It aims to handle more everyday work efficiently, while standard API token prices are half those of GPT-6.1 Sol.
Working across applications
The main improvement highlighted in the announcement is computer use: interpreting visual information and carrying out tasks in software. Examples in OpenAI’s evaluation include turning architectural drawings into 3D models and reviewing animated footage. The company presents these results as benchmark measurements, rather than a guarantee of performance on every task.
On Agents’ Last Exam, OpenAI says Terra at medium reasoning effort performs on par with GPT-6 Astra at one-tenth its price per token. At maximum effort, OpenAI reports that Terra uses more than three times fewer tokens than Claude Opus 5.5 and costs about nine times less per task in that evaluation. That is a benchmark-specific estimate; it should not be taken to mean Terra is universally better or dozens of times cheaper on every job.
OpenAI also says Terra improves on the previous Terra model in visual reasoning, computer use and coding. For particularly complex coding, the company still recommends GPT-6.1 Sol or GPT-6 Astra. Performance in architecture, game development or 3D work will depend on the task and tools used.
Price and availability
Standard API rates are $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens. The standard input and output rates for GPT-6.1 Sol are $2 and $10 respectively. Long-context requests, cache writes, processing mode and tool fees can change the actual cost; subscription usage is separate from API billing.
OpenAI says GPT-6.1 Terra is available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and via the API as gpt-6.1-terra. See the OpenAI announcement, API model page and the original OpenAI Community post for details.
Translated from the original Chinese article, published October 26, 2026. Pricing and benchmark context checked against OpenAI’s announcement.
Comments NOTHING