OpenAI has launched GPT-6.1 Terra, a mid-tier reasoning model designed to make capable AI more affordable for high-volume business and developer workloads. It replaces GPT-5.6 Terra and sits between the lower-cost GPT-6 Luna and OpenAI’s more capable GPT-6.1 Sol and GPT-6 Astra models.openai
Built for everyday professional tasks
OpenAI describes Terra as a fast, efficient model that combines visual reasoning, computer use, and clearer writing. The company says it is intended for the kinds of tasks that fill the workday, such as drafting proposals, interpreting technical drawings, filling out forms, and editing videos.openai
The model is also aimed at automation at scale. OpenAI highlights use cases such as reading incoming invoices and entering their details into accounting software—work that may not require the most expensive frontier model but still needs reliable accuracy and tool use.openai
Lower price, higher efficiency
Terra’s standard API pricing is 50% lower than both GPT-5.6 Terra and GPT-6.1 Sol. Developers pay $1 per million input tokens, $5 per million output tokens, and $0.05 per million cached input tokens.openai
That pricing places Terra below GPT-6.1 Sol, which costs $2 per million input tokens and $10 per million output tokens, and far below GPT-6 Astra, which costs $10 and $50 respectively. GPT-6 Luna remains OpenAI’s lowest-cost option at $0.10 for input and $0.50 for output per million tokens.openai
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
Computer use is the headline capability
OpenAI is positioning computer use as a central strength of Terra. On Agents’ Last Exam, a benchmark for professional tasks, 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
OpenAI says Terra is more than three times as token-efficient as Claude Opus 5.5 at maximum effort and roughly one-quarter of its price per token, resulting in approximately nine times lower cost per task. The company also says Terra completes tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol.openai
Visual reasoning improves substantially
On MMMU-Pro, a visual reasoning benchmark, Terra scored 85.5% at maximum effort at a cost of $0.012 per task. That is an improvement over GPT-5.6 Terra’s 81.7% score at $0.079 per task.openai
Terra comes close to GPT-6.1 Sol, which scored 86.3% at $0.013 per task, and GPT-6 Astra, which led at 86.7% at $0.065 per task. GPT-6 Luna scored 79.9% at $0.0021 per task.openai
On OSWorld 2.1, which measures how well AI agents navigate desktop applications to complete multi-step tasks, Terra scored 74.2% at $1.19 per task. That is a major improvement over GPT-5.6 Terra’s 57.1% score at $4.76 per task, though it still trails Claude Sonnet 5.5’s 83.9% score at $5.74 per task.openai
Fewer factual errors
OpenAI says it invested heavily in reducing hallucinations and factual errors in Terra. 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.openai
At maximum reasoning effort, the error rate dropped to 4.3%, nearly matching GPT-6 Astra and GPT-6.1 Sol. OpenAI says Terra is also more likely to stay within scope, disclose limitations, and resist prompt-injection attempts than its predecessor.openai
Coding gains, with limits
Terra also improves software-engineering performance. On Terminal-Bench 4.0, which tests coding agents solving command-line problems, Terra solved 50.3% of tasks at maximum reasoning effort, more than double GPT-5.6 Terra’s published score of 21.5%.openai
However, OpenAI says GPT-6.1 Sol and GPT-6 Astra remain better choices for the most complex coding tasks. Terra is positioned as the more economical option for routine coding and high-volume automation rather than frontier-level engineering work.openai
The real cost question
Lower token prices improve one part of AI economics, 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 inference, especially when agents operate at scale.openai
For enterprises, the key metric is not price per token but cost per completed task. That includes task completion rates, accuracy, execution time, human intervention, compliance requirements, and the cost of integrating the model into existing systems.infoworld
Safety and governance
Terra meets the Critical threshold for cybersecurity under OpenAI’s Preparedness Framework and retains safeguards developed for GPT-6 Astra. The company says the model includes alignment improvements from the GPT-6 series and showed strong resistance to malicious prompt-injection instructions in its evaluations.openai
As AI systems increasingly take actions rather than simply 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 operations.
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 GPT-5.6 Terra, particularly in computer use, visual reasoning, and coding, at a much lower cost.openai
Still, benchmark results are only a starting point. Organizations should test Terra on their own workloads and determine whether its lower token prices translate into lower total cost per completed task.
