AI · SEMICONDUCTOR · FACT CHECK
SKD WIRETrue
OpenAI has unveiled GPT-6.1 Terra, a new model release the company says lowers the cost of performing tasks by half compared with previous versions. The announcement was accompanied by official image assets published through OpenAI's own distribution channels, including a model art card and deployment-safety evaluation materials.
The new model, referred to as Terra, arrives under the GPT-6.1 designation and is positioned by OpenAI as a cost-efficiency update rather than a headline capability jump. According to the information accompanying the release, the central claim is a 50 percent reduction in the cost of carrying out work tasks, a metric that applies to how much computing and expense is required for a given unit of work.
OpenAI published supporting materials alongside the announcement, including an official art card hosted on the company's content delivery network and evaluation-set images made available through its deployment-safety portal. The presence of dedicated evaluation assets suggests the release is tied to the company's standard safety and performance assessment pipeline, though OpenAI has not detailed the full scope of those evaluations in the materials reviewed for this report.
The "half the cost" framing is the most concrete figure attached to the release. Lowering the cost per task is a recurring theme among frontier-lab updates, as it determines how economically the model can be deployed at scale for routine workloads such as drafting, coding assistance, and document processing.
Details on how the 50 percent figure was measured — whether against GPT-6.1's prior configuration, an earlier generation, or a specific benchmark of task types — were not spelled out in the announcement materials. The company has also not published pricing tables in the assets associated with the release, so the practical impact on API customers will depend on the final rate card that accompanies general availability.
Several elements of the launch remain unconfirmed. The precise relationship between Terra and other models in the GPT-6.1 family, the availability timeline across OpenAI's API and consumer products, and regional rollout plans are not specified in the materials made public so far. The deployment-safety assets indicate the model has passed through OpenAI's internal evaluation process, but the underlying results of those evaluations have not been released in full.
For now, the verifiable core of the announcement stands on two points: OpenAI has publicly released a model it calls GPT-6.1 Terra, and the company presents it as halving the cost of task execution. The claim that GPT-6.1 Terra has been released with task costs reduced by half is True.
