OpenAI launches GPT-6.1 Terra for faster desktop automation

A new model for enterprise tasks

OpenAI has officially introduced GPT-6.1 Terra, a model specifically engineered to handle common business operations. The new system is designed to perform tasks such as drafting documents, conducting visual analysis, and writing code. It also focuses on direct interaction with desktop applications to automate workflows.

The company aims to provide a more cost-effective solution for high-volume tasks that do not require the highest level of reasoning. By optimizing for these routine actions, GPT-6.1 Terra serves as a practical tool for software developers and large organizations. It is now available to Plus, Pro, Enterprise, and Edu subscribers.

As reported by LatestLY, the model is also available through standard API access. To encourage adoption for large-scale automation, OpenAI has cut the standard API pricing by 50 percent compared to previous versions. This move makes it easier for companies to integrate the model into their existing software stacks.

Benchmark gains and speed

The new model shows significant improvements over its predecessor, GPT-5.6 Terra. On the Agents' Last Exam benchmark, it scored 15 percentage points higher at a medium reasoning effort. This improvement comes while the model operates at roughly one-third of the operational cost of the previous version.

Speed is another primary focus for this release. GPT-6.1 Terra can complete tasks more than three times faster than higher-tier models like GPT-6 Astra and GPT-6.1 Sol. These performance gains are intended to help users who need rapid responses for repetitive computer-use evaluations.

When tested on the OSWorld 2.1 evaluation for navigating desktop applications, the model showed high token efficiency. It achieved up to a ninefold cost reduction per task when compared to other models currently in the industry. These metrics suggest that the model is built to be both fast and economical for heavy-duty usage.

Accuracy and safety improvements

OpenAI has made notable progress in reducing hallucinations, which are instances where an AI generates false information. Internal tests showed that the factual error rate at low reasoning effort dropped from 53.6 percent in the previous generation to 16.7 percent. At maximum reasoning effort, the error rate was reduced to 4.3 percent.

The model also demonstrated specific capabilities in technical environments. On the Terminal-Bench 4.0 coding tasks, it successfully solved 50.3 percent of the command-line challenges presented. This indicates a growing proficiency in handling structured, technical instructions.

Security remains a core component of the release. The model retains the safety protocols established under the Preparedness Framework. It features improved resistance against prompt-injection attacks, which are attempts to bypass safety filters, and it shows better adherence to defined task boundaries.

Who can use the new model

The launch of GPT-6.1 Terra provides immediate access to a wide range of users. Subscribers to ChatGPT Work, Codex, and the various paid tiers can begin using the model today. This includes users in the education sector who require reliable tools for drafting and analysis.

For developers, the 50 percent price cut on the API is a major update for building automated systems. Because the model is optimized for desktop navigation, it is particularly useful for creating agents that can click buttons or navigate menus. This makes it a viable option for companies looking to automate manual office tasks.

The release follows a trend of OpenAI refining its models to be more specialized. While previous models might have been general-purpose, GPT-6.1 Terra is tailored for the specific needs of enterprise environments. It balances the need for high performance with the necessity of lower operational costs for large-scale deployment.

The path forward for automation

The introduction of GPT-6.1 Terra highlights a shift toward models that can interact directly with computer interfaces. By focusing on desktop application navigation, OpenAI is moving toward a future where AI can act as a digital assistant. This requires the model to understand visual cues and software layouts in real-time.

The significant reduction in cost per task is a key factor for the next wave of AI adoption. As businesses look to deploy AI agents, the ability to run these models at a fraction of the previous cost is essential. This helps move AI from a novelty to a standard part of the corporate infrastructure.

Moving forward, users can expect to see more integration of these capabilities into existing software. With the model's improved accuracy and lower error rates, it becomes a more reliable choice for sensitive tasks. The focus remains on making high-volume workflow automation accessible to a broader range of software developers.