OpenAI releases GPT 6.1 Terra

OpenAI has introduced GPT-6.1 Terra, a new model for working with text, images, and computer interfaces. It approaches the capabilities of the family's larger models at half the price of GPT-6.1 Sol. Its main focus is computer use: tasks that require a model to navigate an interface and carry out a sequence of actions, rather than simply answer a question.
Terra can read technical drawings, match product photos to catalog entries, fill out forms, and work with applications. OpenAI also reports more natural writing and fewer factual errors.

Benchmark performance

On Agents' Last Exam, which evaluates professional tasks, GPT-6.1 Terra at medium reasoning effort outperforms GPT-5.6 Terra by 15 percentage points, at nearly one-third of the cost per task.At maximum reasoning effort, OpenAI reports that the new model completes tasks more than three times t as GPT-6 Astra and GPT-6.1 Sol. On MMMU-Pro, a visual understanding benchmark, Terra comes close to the family's larger models.Coding performance has also improved: its Terminal-Bench 4.0 score rose to 50.3%, up from 21.5% for the previous version.

Pricing

Input: $1 per million tokens.
Output: $5 per million tokens.
Cached input: $0.05 per million tokens.

For comparison, GPT-6.1 Sol costs $2 per million input tokens and $10 per million output tokens.Terra has a 1.05-million-token context window and supports up to 128,000 output tokens. Multiple reasoning levels are available, from none to max.

Availability

The model is already available through the API and is rolling out in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users. It will not be available in the standard ChatGPT chat interface. It is also available through BotHub.