OpenAI has launched GPT-6.1 Terra, a mid-tier reasoning model designed to deliver strong performance on visual reasoning, professional tasks, and computer use at a substantially lower cost than its frontier models. It replaces GPT-5.6 Terra and is positioned for enterprises that want to automate high-volume work without paying premium prices.openai

A model for everyday professional work

Terra sits between OpenAI’s lower-cost GPT-6 Luna and its more capable GPT-6.1 Sol and GPT-6 Astra models. OpenAI describes it as a fast, efficient option for tasks that fill much of the working day, including drafting proposals, interpreting technical drawings, filling out forms, and editing videos.openai

The company says Terra is a significant improvement over GPT-5.6 Terra, with standard API pricing 50% lower. That makes it more practical for repetitive, high-volume work such as reading incoming invoices and entering their details into accounting systems.openai

Stronger computer-use performance

OpenAI is emphasizing Terra’s ability to operate software interfaces and complete multi-step tasks. On Agents’ Last Exam, a benchmark for professional work, Terra scored 60.0% at maximum reasoning effort at a cost of $1.01 per task.openai

That compares with 50.7% for GPT-5.6 Terra at $3.58 per task, 58.6% for GPT-6.1 Sol at $2.00, and 59.3% for GPT-6 Astra at $7.24. Claude Opus 5.5 scored 63.2% at $8.82 per task.openai

On OSWorld 2.1, which tests how AI agents navigate desktop applications, Terra scored 74.2% at $1.19 per task, up from GPT-5.6 Terra’s 57.1% at $4.76. It still trails Claude Sonnet 5.5, which led the benchmark at 83.9% with a cost of $5.74 per task.openai

Visual reasoning and accuracy

On the MMMU-Pro visual reasoning benchmark, Terra scored 85.5% at maximum effort, costing $0.012 per task. That is an improvement over GPT-5.6 Terra’s 81.7% score at $0.079 per task and close to GPT-6.1 Sol’s 86.3% and GPT-6 Astra’s 86.7%.openai

OpenAI also says Terra significantly reduces factual errors. On the company’s internal benchmark of error-inducing prompts, the error rate at low reasoning effort fell from 53.6% with GPT-5.6 Terra to 16.7% with Terra. At maximum effort, the error rate fell to 4.3%, nearly matching GPT-6 Astra and GPT-6.1 Sol.openai

Pricing and availability

GPT-6.1 Terra is available to ChatGPT Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex. Developers can access it through the OpenAI API as gpt-6.1-terra.openai

API pricing is:

• $1 per million input tokens.
• $5 per million output tokens.
• $0.05 per million cached input tokens.

By comparison, GPT-6.1 Sol costs $2 per million input tokens and $10 per million output tokens, while GPT-6 Astra costs $10 and $50 respectively. GPT-6 Luna remains the lowest-cost option at $0.10 for input and $0.50 for output per million tokens.openai

Where Terra fits

Terra is designed for high-volume work such as invoice processing, document analysis, application automation, and routine coding. For more complex reasoning, research, and software engineering, OpenAI says GPT-6.1 Sol and GPT-6 Astra remain better choices.openai

In coding tests, Terra solved 50.3% of Terminal-Bench 4.0 tasks at maximum reasoning effort, more than double GPT-5.6 Terra’s published score of 21.5%. Still, OpenAI says its higher-tier models are better suited to the most difficult coding work.openai

The economics are not just token prices

Lower inference costs improve one part of the AI cost equation, but they do not eliminate the cost of reviewing, integrating, and correcting AI output. Analysts note that even small increases in human review time can erase savings from cheaper tokens, especially at scale.openai

For enterprises, the practical question is not simply which model is cheapest per token, but which model delivers completed work that is accurate, authorized, and ready for use. That requires testing on real workflows and measuring task completion, accuracy, execution time, cost per workflow, and the level of human intervention required.infoworld

Safety and governance

OpenAI says Terra meets the Critical threshold for cybersecurity under its Preparedness Framework and retains safeguards developed for GPT-6 Astra. The company also says Terra is more likely to stay within scope, disclose limitations, and resist prompt-injection attempts than GPT-5.6 Terra.openai

As AI models increasingly take actions rather than merely generate text, governance becomes more important. Enterprises deploying computer-use agents should maintain clear permission boundaries, audit trails, and human approval controls for sensitive or irreversible actions.infoworld

The takeaway

GPT-6.1 Terra reflects OpenAI’s effort to make capable AI more economically viable for routine enterprise work. Its benchmark results suggest meaningful gains over its predecessor, particularly in computer use and visual reasoning, at a much lower cost.

But benchmark performance is only a starting point. Organizations should validate Terra against their own workloads and determine whether its lower token prices translate into lower total cost per completed task.