GPT-6.1 Terra promises faster AI agents at a fraction of the cost

OpenAI has introduced GPT-6.1 Terra, a new model designed for visual reasoning, computer use, writing, and coding tasks at lower prices than its predecessor. OpenAI announces in an official blog post that the model is available in ChatGPT Work, Codex, and through its API.

The company positions Terra as a model for high-volume workplace tasks. Examples include reading invoices, entering data into accounting software, interpreting technical drawings, completing forms, and editing video. OpenAI says its standard API pricing is 50% lower than GPT-5.6 Terra and GPT-6.1 Sol.

Developers pay $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens. Tokens are units of text that AI systems process. The lower rates could matter most for teams that run many repetitive tasks through automated workflows.

Focus on visual and computer-based work

According to OpenAI, GPT-6.1 Terra improves on GPT-5.6 Terra in benchmarks that test the use of desktop software and visual information. On Agents’ Last Exam, which includes tasks such as converting architectural drawings into 3D models, Terra scored 15 percentage points higher at medium reasoning effort while costing about one-third as much, the company says.

OpenAI also reports that the model completes tested tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol. On the visual reasoning benchmark MMMU-Pro, Terra nearly reaches the results of those more capable models at a lower cost, according to the company.

Fewer errors and stronger coding results

OpenAI says Terra produces fewer factual errors on its internal test set. Its reported error rate at low reasoning effort fell from 53.6% for GPT-5.6 Terra to 16.7% for the new model. At maximum effort, the company reports a 4.3% error rate.

The model also more than doubled GPT-5.6 Terra’s published result on Terminal-Bench 4.0, a command-line coding benchmark. OpenAI notes that GPT-6.1 Sol and GPT-6 Astra remain better suited to the most complex programming work. The company says Terra retains the GPT-6 series’ cybersecurity safeguards and improved resistance to prompt injection.