OpenAI has launched GPT-6.1 Terra, a new "fast and efficient" model that combines detailed visual reasoning with precision in computer use and clearer text output. The model is presented as a step up from GPT-5.6 Terra and targets the tasks that fill a normal working day — everything from drafting proposals and reading technical drawings to filling in forms and editing video.

Lower cost, more work per license

The central argument is pricing. In OpenAI's standard API, GPT-6.1 Terra is 50 percent cheaper than both GPT-5.6 Terra and GPT-6.1 Sol. In ChatGPT Work and Codex, the same usage quota covers more tasks with Terra than with GPT-5.6 Terra, and significantly more than with GPT-6.1 Sol or GPT-6 Astra.

Computer use and visual reasoning

On Agents' Last Exam, which measures professional tasks, GPT-6.1 Terra scores 15 percentage points higher than GPT-5.6 Terra at medium reasoning for roughly a third of the cost. At maximum effort, Terra is more than three times as token-efficient as Claude Opus 5.5 and costs a quarter of its per-token price, giving a roughly nine times lower cost per task. Terra is also faster — tasks are completed more than three times faster than with both GPT-6 Astra and GPT-6.1 Sol.

On MMMU-Pro, a visual-reasoning benchmark, GPT-6.1 Terra beats GPT-5.6 Terra at all reasoning settings and approaches the most capable models, GPT-6 Astra and GPT-6.1 Sol. On OSWorld 2.1, which tests agents using desktop applications in multi-step tasks, Terra outperforms GPT-5.6 Terra's best results at medium reasoning for a fraction of the cost per task.

Factuality and coding

OpenAI highlights a major push to reduce hallucinations and factual errors. On an internal benchmark with misleading prompts, the error rate at low reasoning fell from 53.6 percent with GPT-5.6 Terra to 16.7 percent with GPT-6.1 Terra — a reduction of nearly 37 percentage points. At maximum effort, the error rate dropped 18 percentage points to 4.3 percent, in line with GPT-6 Astra and GPT-6.1 Sol.

On Terminal-Bench 4.0, which tests coding agents through the command line, GPT-6.1 Terra solves 50.3 percent of tasks at maximum reasoning — more than double GPT-5.6 Terra's published result of 21.5 percent. For the most complex coding tasks, OpenAI states that GPT-6.1 Sol and GPT-6 Astra remain the preferred choice.

Safety and availability

According to OpenAI, GPT-6.1 Terra meets the "Critical" threshold for cybersecurity under the company's Preparedness Framework and retains the protections developed for GPT-6 Astra. In the company's evaluations, the model is more likely to stay within scope and honestly acknowledge its limitations compared with GPT-5.6 Terra, and it showed strong resistance to malicious instructions in prompt-injection tests.

The model is available from today for Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and via the OpenAI API under the name gpt-6.1-terra.

Why it matters

GPT-6.1 Terra is positioned in OpenAI's model family as the cost-effective counterpoint to the most advanced frontier models, GPT-6 Astra and GPT-6.1 Sol. The release continues a clear trend of moving the best results to cheaper, faster models rather than keeping them in the top model. For developers, it means high-volume tasks — such as reading incoming invoices and entering them into accounting systems — become significantly cheaper to automate. More technical detail is in OpenAI's system card.

Sources: