Quick Answer
GPT-6.1 Terra is OpenAI’s new model for visual reasoning, precise computer use, and natural writing, available in ChatGPT Work and Codex for eligible paid users and through the API. Its standard API price starts at $1 per million input tokens, which OpenAI says is 50% below GPT-5.6 Terra and GPT-6.1 Sol. Developers should still evaluate outputs and safeguards before deploying computer-use workflows.
Key Takeaways
- GPT-6.1 Terra launched on October 26, 2026.
- GPT-6.1 Terra supports visual reasoning, computer use, and natural writing.
- Standard API input pricing is $1 per million tokens.
- The model has a 1.05-million-token context window and 128,000-token output limit.
- OpenAI classifies GPT-6.1 Terra as Critical for cybersecurity capability.
What is GPT-6.1 Terra?
GPT-6.1 Terra is an OpenAI model designed for visual reasoning, precise computer use, and natural writing. OpenAI announced the model on October 26, 2026, and positioned it for tasks where an AI system needs to interpret visual information, work through software interfaces, and produce readable responses. OpenAI’s launch announcement describes Terra as a model intended to combine those capabilities in one system.
GPT-6.1 Terra matters because computer-use systems are meant to take actions through an interface rather than only return text. A workflow could require the model to inspect a page, identify the relevant controls, and continue through a defined task. The practical limitation is that a model’s ability to interact with a computer does not remove the need for human oversight, especially when an action could affect accounts, files, purchases, or published material.
GPT-6.1 Terra also arrives as AI companies are expanding agent-style products that can operate across connected services. Users considering tools with broad permissions should review the same issues raised by AI access to calendars and contacts: a useful feature can also expose more personal information when its permissions are broader than the task requires.
Who can use GPT-6.1 Terra?
GPT-6.1 Terra is available in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users, and developers can access the model through the API under the name gpt-6.1-terra. OpenAI announced those availability details on October 26, 2026. The split matters because ChatGPT access and API access serve different needs: one is intended for product users, while the other lets developers build Terra into their own applications and workflows.
GPT-6.1 Terra availability does not mean every user receives the same tools, controls, or administrative setup. Enterprise and Edu organizations can have their own access policies, while developers must decide which functions to enable in an API integration. The most sensible approach is to confirm the model option in the relevant ChatGPT or Codex workspace before changing an existing workflow.
GPT-6.1 Terra also enters a market where model access is increasingly distributed through cloud services and third-party software. OpenAI has already expanded model availability through Google Cloud model access, but availability through a platform does not automatically establish the same feature set or operational controls as OpenAI’s direct products.
How much does GPT-6.1 Terra cost in the API?
GPT-6.1 Terra costs $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens at OpenAI’s standard API rates. OpenAI also says the standard pricing is 50% lower than GPT-5.6 Terra and GPT-6.1 Sol. OpenAI’s model documentation lists the pricing and model configuration details.
GPT-6.1 Terra’s cached-input price matters for applications that repeatedly send the same context, such as stable instructions or large reference materials. Lower input costs can reduce the expense of repeated requests, but output remains priced separately, so a workflow that generates long responses can still carry meaningful usage costs. Developers should estimate both input and output volume before treating the listed input rate as the total cost of an application.
| GPT-6.1 Terra API item | OpenAI-listed figure | Why it matters |
|---|---|---|
| Input tokens | $1 per million | Applies to the content sent to the model. |
| Cached input tokens | $0.05 per million | Can reduce costs when applications reuse context. |
| Output tokens | $5 per million | Applies to the content generated by the model. |
| Price comparison | 50% lower than GPT-5.6 Terra and GPT-6.1 Sol | OpenAI’s comparison applies to standard API pricing. |
GPT-6.1 Terra pricing is an API measure, not a published flat price for every ChatGPT plan or enterprise agreement. The practical response is to separate model usage costs from subscription access costs and confirm current billing details in the account or contract that applies to your organization.
What are GPT-6.1 Terra’s technical limits and tools?
GPT-6.1 Terra has a 1.05-million-token context window and a maximum output of 128,000 tokens, according to OpenAI’s API documentation. The model also supports functions, web search, file search, and computer use. Those specifications give developers room to provide substantial context and connect the model to external tools, but they do not guarantee that every response or tool action will be correct.
GPT-6.1 Terra’s long context window is most relevant when an application needs to keep extensive instructions, documents, or prior material available in a single request. A larger context window can reduce the need to split source material into many smaller prompts. At the same time, developers still need to decide what information belongs in a request, because sending unnecessary sensitive data creates avoidable privacy and security exposure.
GPT-6.1 Terra’s computer-use support also makes permission design important. Developers should limit the accounts, browser sessions, folders, and tools available to an automated workflow. The same principle applies to browser-based assistants that can act on open pages, including the broader concerns around AI control of Chrome tabs.
How does GPT-6.1 Terra perform on OpenAI’s tests?
GPT-6.1 Terra recorded a 16.7% factual-error rate at low reasoning effort on OpenAI’s internal error-inducing prompt benchmark, compared with 53.6% for GPT-5.6 Terra, according to OpenAI. The company also reports that Terra solved 50.3% of Terminal-Bench 4.0 software-engineering tasks at maximum reasoning effort, while GPT-5.6 Terra’s previously published result was 21.5%. These are OpenAI-reported benchmark results, and independent testing may produce different outcomes.
GPT-6.1 Terra’s reported scores suggest that OpenAI sees the model as an improvement for factual reliability and software-engineering tasks under the tested conditions. The limitation is that benchmark performance measures a defined evaluation, not every real-world task a person or business may assign to the model. A model can perform well on a software task benchmark and still make an unsuitable decision when a task has incomplete instructions, unusual data, or consequences outside the benchmark.
GPT-6.1 Terra should therefore be tested against the actual documents, software, and approval requirements of a planned deployment. Developers should use representative tasks, review incorrect actions, and maintain a human approval step for high-impact work rather than treating a benchmark result as proof of dependable autonomy.
What safety classification does GPT-6.1 Terra have?
GPT-6.1 Terra is classified by OpenAI as Critical for cybersecurity capability and High for biological and chemical capability under the company’s Preparedness Framework. OpenAI says it applies the safeguards stack used for GPT-6.1 Sol and GPT-6 Astra to Terra. OpenAI’s GPT-6.1 Terra safety addendum sets out those classifications and the company’s stated safety approach.
GPT-6.1 Terra’s cybersecurity classification matters because a model that is stronger at computer interaction and technical reasoning can create greater misuse concerns alongside legitimate benefits. The classification is OpenAI’s assessment under its own framework, not a guarantee that the safeguards eliminate all harmful use or that every deployment carries the same level of risk.
GPT-6.1 Terra users should keep the model away from systems where an unreviewed action could expose credentials, modify security controls, or distribute code. Organizations handling sensitive environments should involve their security team before enabling computer use or tool access. Stop automated testing and contact the relevant security owner if a workflow begins accessing data or taking actions outside its intended permissions.
How should developers use GPT-6.1 Terra safely?
GPT-6.1 Terra should be deployed with narrow permissions, clear task boundaries, and human review for consequential actions. OpenAI supports functions, web search, file search, and computer use for the model, which can make an application more useful but also increases the number of systems and data sources a workflow may reach. The practical safety decision is to expose only the tools and information required for a specific task.
- Define the allowed task before connecting GPT-6.1 Terra to a tool or computer environment.
- Limit access to the minimum folders, accounts, websites, and functions needed for that task.
- Require human approval before sending messages, changing records, making purchases, or publishing content.
- Test GPT-6.1 Terra with representative inputs and document the errors that require a fallback process.
- Review logs and stop the workflow if the model attempts actions outside the approved scope.
GPT-6.1 Terra should not receive passwords, payment details, medical records, or other sensitive information unless an organization has determined that the data handling and permissions are appropriate. The risks of overly broad access have become more visible as AI tools gain persistent context and connected services, including personal data access in cars. Contact the vendor or an internal security professional before deploying a computer-use workflow in a production environment with sensitive systems.
FAQ
What is GPT-6.1 Terra used for?
GPT-6.1 Terra is used for visual reasoning, precise computer use, natural writing, and API-based applications that need functions, web search, file search, or computer-use support. OpenAI announced the model on October 26, 2026.
How much does GPT-6.1 Terra cost?
GPT-6.1 Terra costs $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens at OpenAI’s standard API rates. OpenAI says those standard prices are 50% lower than GPT-5.6 Terra and GPT-6.1 Sol.
Who can access GPT-6.1 Terra?
GPT-6.1 Terra is available in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users. Developers can also call the model through the API as gpt-6.1-terra.
What is GPT-6.1 Terra’s context window?
GPT-6.1 Terra has a 1.05-million-token context window and a maximum output of 128,000 tokens. The large context limit can support extensive source material, but developers should still avoid sending unnecessary sensitive information.
Is GPT-6.1 Terra safe for computer-use tasks?
GPT-6.1 Terra can support computer-use tasks, but users should restrict permissions and require human approval for consequential actions. OpenAI classifies the model as Critical for cybersecurity capability, so production deployments need appropriate security review and monitoring.