OpenAI launched GPT-6.1 Terra on Monday, October 26, 2026, positioning the efficient mid-tier model as a practical upgrade for high-volume professional work. The release combines detailed visual reasoning, precise computer use and clearer natural writing while cutting standard API prices by half. GPT-6.1 Terra is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and to developers through the API as gpt-6.1-terra.
The timing matters because Terra was absent when OpenAI updated the higher Sol and lower Luna tiers last month. GPT-6.1 Terra is therefore the medium option in the GPT-6 family arriving after the rest of the lineup. Its central pitch is not simply stronger intelligence, although the benchmarks improved. It is the combination of better accuracy, faster task completion and lower unit pricing for organizations running agents across many routine workflows.
What OpenAI is claiming for GPT-6.1 Terra
OpenAI describes GPT-6.1 Terra as a significant step up from GPT-5.6 Terra, particularly for communication and computer-based tasks that occupy much of the working day. Typical examples include drafting proposals, reading technical drawings, completing forms and editing video. Those activities require a model to understand what appears on a screen, choose the right interface element, carry out a sequence of actions and produce a clear account of the result.
Standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens. That is 50 percent below both GPT-5.6 Terra and GPT-6.1 Sol. The price reduction applies to the standard rates, so developers can compare operating costs more directly when deciding whether to move high-volume workloads to the new model.
On OpenAI’s Agents’ Last Exam benchmark of professional computer-use tasks, GPT-6.1 Terra scored 15 percentage points higher than GPT-5.6 Terra at roughly one third of the cost. It also performed on a par with GPT-6 Astra at one tenth of Astra’s price per token. At maximum reasoning effort, Terra used more than three times as many tokens efficiently as Claude Opus 5.5 at one quarter its price per token, which OpenAI translates into a roughly nine-fold lower cost per task. The company also says GPT-6.1 Terra completed tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol.
Accuracy gains beyond raw speed
The speed claim is paired with a substantial reduction in errors. On an internal benchmark built from prompts designed to induce mistakes, the error rate at low reasoning effort fell from 53.6 percent with GPT-5.6 Terra to 16.7 percent with GPT-6.1 Terra. At maximum reasoning effort, the rate fell by 18 percentage points to 4.3 percent, nearly matching GPT-6 Astra and GPT-6.1 Sol.
Visual reasoning also improved across every reasoning setting on MMMU-Pro, where GPT-6.1 Terra beat GPT-5.6 Terra and nearly matched GPT-6 Astra. On OSWorld 2.1, a test of desktop agents, it surpassed the previous Terra’s best score at a fraction of the cost per task. These results suggest that the efficiency gains do not come from simply asking the model to take shorter paths. OpenAI is presenting Terra as a stronger tool that happens to require less time and money to complete the job.
Coding results are another significant part of the release. On Terminal-Bench 4.0, GPT-6.1 Terra solved 50.3 percent of tasks at maximum reasoning effort, more than double GPT-5.6 Terra’s published score of 21.5 percent. On Terminal-Bench-Science 0.1, it surpassed 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. For teams building coding agents, the new model narrows the distance between the everyday tier and more expensive frontier systems.
Where developers and teams can use it
Access is broad from the first day. GPT-6.1 Terra is offered through the API as gpt-6.1-terra and appears in ChatGPT Work, OpenAI’s agent mode for working across applications and completing tasks on websites. It is also included in Codex for coding work. Plus, Pro, Enterprise and Edu users can access it, while Free users remain on the lower-cost GPT-5.6 Luna tier according to reporting on the release.
Amazon Bedrock also announced general availability on the same day with a context window of up to one million input tokens. GitHub Copilot added the model for Pro, Pro+, Max, Business and Enterprise customers. That distribution gives developers more than one route for testing the model and makes comparison with other hosted models easier.
OpenAI continues to steer demanding research and complex coding toward GPT-6.1 Sol or GPT-6 Astra. Terra is positioned for workloads where users need many completed actions rather than a single exceptional answer, such as high-volume form processing, document analysis and repetitive desktop automation. The long context window on Bedrock may also help with repositories, archives and other large collections of work material.
Safety classification and limitations of the evidence
Under OpenAI’s Preparedness Framework, GPT-6.1 Terra is classified as Critical in cybersecurity and High for biological and chemical capability. The model uses the same safeguards stack as GPT-6.1 Sol and GPT-6 Astra, with probes updated for compatibility. It also carries the alignment improvements associated with the GPT-6 series. On Production Benchmarks, OpenAI reports performance broadly on par with GPT-5.6 Terra, with no statistically significant regressions in extremism, hate or sexual content involving minors.
Most of the headline performance figures come from OpenAI’s own evaluations, and the system card notes that results from its research environment or API may differ from production ChatGPT because of differences in system prompts, tools and reasoning effort. Benchmark scores also do not capture every factor in production. Reliability against unfamiliar interfaces, latency under heavy load, rate limits and the cost of corrective human review will determine how useful Terra proves to be in practice. That is the trade GPT-6.1 Terra is priced on, and the next round of enterprise renewals will show whether a 50 percent price cut is enough to move high-volume workloads onto a mid-tier frontier model.
Even with those caveats, GPT-6.1 Terra is a notable release for buyers who need dependable agents at scale. It combines a large error reduction, faster execution, improved visual understanding and stronger coding results with a 50 percent price cut. With immediate availability across ChatGPT Work, Codex, the API, Amazon Bedrock and GitHub Copilot, GPT-6.1 Terra gives organizations a practical middle tier for automating everyday professional work.