OpenAI launches new GPT-6.1 Terra for workplace tasks

OpenAI has launched GPT-6.1 Terra, now available in ChatGPT Work, Codex and the OpenAI API.

OpenAI says GPT-6.1 Terra is designed for workplace tasks that combine writing, visual reasoning and computer use. It is positioned as the successor to GPT-5.6 Terra, with lower pricing and stronger performance on a range of internal and external tests.

Standard API pricing is USD $1 per million input tokens, USD $0.05 per million cached input tokens and USD $5 per million output tokens. OpenAI says that is 50% lower than the standard API prices for GPT-5.6 Terra and GPT-6.1 Sol.

The model is intended for high-volume administrative and software tasks, including reading invoices and entering details into accounting software, drafting proposals, interpreting technical drawings, filling out forms and editing videos. It is also available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex.

Benchmark results

OpenAI published several benchmark comparisons showing how GPT-6.1 Terra performs against earlier models and rival systems. On Agents' Last Exam, which tests professional computer-based tasks, GPT-6.1 Terra scored 15 percentage points higher than GPT-5.6 Terra at medium reasoning effort while running at about one-third of the cost, according to the company.

OpenAI also says GPT-6.1 Terra performed on a par with GPT-6 Astra on that benchmark at one-tenth of Astra's price per token. At maximum reasoning effort, the model was more than three times as token-efficient as Claude Opus 5.5 and around one-quarter of its price per token, producing what OpenAI described as about nine times lower cost per task.

Speed was another point of comparison. OpenAI says GPT-6.1 Terra completed tasks more than three times faster than both GPT-6 Astra and GPT-6.1 Sol.

On MMMU-Pro, a visual reasoning benchmark, GPT-6.1 Terra outperformed GPT-5.6 Terra at every reasoning setting, according to OpenAI. The company says the model nearly matched GPT-6 Astra and GPT-6.1 Sol on scores while operating at lower cost.

Results on OSWorld 2.1, which measures how AI agents use desktop applications to complete multistep tasks, also showed gains over the earlier Terra model. OpenAI says GPT-6.1 Terra exceeded GPT-5.6 Terra's best score at medium reasoning effort at a fraction of the cost per task.

Factuality and coding

OpenAI says it has focused this year on reducing hallucinations and factual mistakes. On the company's internal benchmark of difficult prompts designed to induce errors, GPT-6.1 Terra reduced the factual error rate at low reasoning effort to 16.7% from 53.6% for GPT-5.6 Terra.

At maximum reasoning effort, the error rate fell to 4.3%, according to OpenAI. The company says that result came close to the rates recorded for GPT-6 Astra and GPT-6.1 Sol.

The new model also showed stronger results on software engineering tests. On Terminal-Bench 4.0, which measures how coding agents solve command-line tasks, GPT-6.1 Terra solved 50.3% of tasks at maximum reasoning effort, compared with GPT-5.6 Terra's published score of 21.5%, OpenAI says.

OpenAI says GPT-6.1 Sol and GPT-6 Astra remain the better option for the most complex coding work. On Terminal-Bench-Science 0.1, which tests how agents use code for scientific research tasks, GPT-6.1 Terra outperformed GPT-6.1 Sol at lower cost and came within five points of Claude Opus 5.5 at less than one-fifth of the estimated cost, according to the company.

Safety measures

OpenAI says GPT-6.1 Terra meets the Critical threshold in cybersecurity under its Preparedness Framework. The model also retains the protections introduced for GPT-6 Astra.

According to OpenAI's evaluations, GPT-6.1 Terra is more likely than GPT-5.6 Terra to stay within the scope of a user's request and to disclose its limitations or blockers honestly. Prompt-injection tests also showed strong resistance to malicious instructions intended to divert the model from the user's task, OpenAI says.

"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."