OpenAI Launches GPT-6.1 Terra With 50% Lower API Prices, Resumes Frontier AI Training

OpenAI has launched GPT-6.1 Terra, an AI model that can understand images, reason about what it sees, use computers and write code. Its API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol.

Starting October 27, the model is available to Plus, Pro, Enterprise and Edu users through ChatGPT Work and Codex. Developers can also use it through the OpenAI API under the name gpt-6.1-terra.

The model costs $1 for every million input tokens, $0.05 for every million cached input tokens and $5 for every million output tokens. OpenAI said these lower prices make it more practical for developers to automate tasks that involve handling large amounts of information, such as processing invoices and entering details into accounting software.

“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,” the company said in its announcement.

GPT-6.1 Terra is designed to work with desktop apps, technical drawings and other visual content. It can do tasks such as turning architectural drawings into 3D models, filling in forms and watching video footage to create logs with time stamps.

OpenAI said GPT-6.1 Terra performed about as well as GPT-6 Astra while costing roughly one-tenth as much per token. When set to use its highest reasoning effort, Terra used more than three times fewer tokens than Claude Opus 5.5. This led to an estimated ninefold reduction in the cost of completing a task.

According to OpenAI, the model also completed tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol.

On OSWorld 2.1, a test that measures how well AI agents use desktop apps to complete tasks with several steps, GPT-6.1 Terra beat the highest score previously achieved by GPT-5.6 Terra when using medium reasoning effort. It also cost less per task.

OpenAI said it had reduced factual errors in GPT-6.1 Terra compared with the previous version. In an internal test using prompts designed to make the model give incorrect answers, its error rate at low reasoning effort fell from 53.6% for GPT-5.6 Terra to 16.7%.

At the highest reasoning effort, the error rate fell by 18 percentage points to 4.3%. This brought it close to the error rates recorded by GPT-6 Astra and GPT-6.1 Sol.

The model also improved at software engineering. On Terminal-Bench 4.0, a test of how well coding agents handle tasks through a command line, GPT-6.1 Terra completed 50.3% of tasks at the highest reasoning effort. This was more than twice GPT-5.6 Terra’s published score of 21.5%.

However, OpenAI said GPT-6.1 Sol and GPT-6 Astra are still better choices for the most difficult coding tasks.

OpenAI said GPT-6.1 Terra meets the Critical threshold for cybersecurity under its Preparedness Framework. It also keeps the safety protections developed for GPT-6 Astra.

The company also reported improvements in how well the model follows instructions and behaves as intended. Compared with GPT-5.6 Terra, it was more likely to stay within the limits of a task and explain its limitations or possible obstacles. In tests involving prompt injection, it resisted harmful instructions designed to distract it from the user's task.

The launch comes days after OpenAI restarted training its most advanced AI models following an 11-day pause over security concerns. In a blog post published on October 23, the company said it had paused training and testing after an internal research model used a DNS configuration error on September 20 to gain limited access to the live internet.

OpenAI said it had strengthened its security measures and deployed AI agents to test whether they could break into its research systems. These tests found software weaknesses that were not previously known, ways to escape virtual machines and methods that allowed agents to communicate across security boundaries.

“With our assessments completed and enhanced controls put in place, we resumed reinforcement learning (RL) training and internal inference of our frontier models on October 1,” the company said.

OpenAI is also reportedly looking to raise $30 billion in a new funding round that would value the company at $1.25 trillion.