A practical launch, not just a bigger model
OpenAI has introduced GPT-6.1 Terra, positioning the model as a fast and efficient step forward for multimodal work, computer use and natural writing rather than as a single “biggest model” release . The announcement frames Terra as a significant upgrade over GPT-5.6 Terra, with particular emphasis on the kinds of tasks that make up professional workflows: interpreting technical drawings, drafting proposals, filling out forms, reviewing visual material and automating repetitive software tasks .
The strategic message is clear: Terra is designed to move the GPT-6 family further down the cost curve without abandoning advanced reasoning. OpenAI says the model combines detailed visual reasoning, precise computer use and clearer writing, while also giving users and developers a cheaper way to run large volumes of work . That makes this launch important not only as a model update, but as a pricing and deployment signal. OpenAI is trying to make stronger agents economical enough for routine business processes, not only premium one-off tasks.
What GPT-6.1 Terra is built to do
The model’s most important positioning is in the middle of a capability-cost trade-off. OpenAI presents GPT-6.1 Terra as more capable than GPT-5.6 Terra and more economical than higher-tier options such as GPT-6 Astra and GPT-6.1 Sol for many tasks . In practice, that means Terra is meant to handle the “long tail” of knowledge-work automation: reading invoices, entering details into business software, matching product photos to catalog entries, interpreting technical drawings, acting on application screens and drafting or revising text .
The OpenAI API model documentation reinforces that use-case profile. It describes GPT-6.1 Terra as suitable for matching product photos to catalog records, interpreting technical drawings, acting on app screens and writing or revising text . Those examples matter because they show the model is not being pitched only as a chat assistant. It is being pitched as an agentic component that can perceive, reason and act across software interfaces.
For developers, the API model ID is gpt-6.1-terra, and OpenAI says tool calling for Terra requires the Responses API . The model documentation also lists a 1,050,000-token context window and a 128,000-token maximum output, with reasoning-token support and a June 1, 2026 knowledge cutoff . The large context window is especially relevant for enterprise and coding workflows, where the practical limit is often not raw intelligence but how much project context, policy text, design material or operational data a model can hold at once.
Visual reasoning and computer use are the center of the story
The most concrete performance claims in OpenAI’s launch post focus on visual reasoning and computer-use benchmarks. OpenAI says GPT-6.1 Terra was measured on Agents’ Last Exam, a benchmark of professional tasks such as converting architectural drawings into a 3D model and reviewing an animated film shot by shot to create a time-coded restoration log . At medium reasoning effort, the company says Terra scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost, and performs on par with GPT-6 Astra at one-tenth the price per token .
OpenAI also says that, at maximum effort, Terra is over three times as token-efficient as Claude Opus 5.5, one-fourth its price per token and roughly nine times cheaper per task, while completing tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol . These are OpenAI’s own benchmark claims, so they should be read as vendor-supplied figures rather than independent validation. Still, they explain the product logic: OpenAI is not simply saying Terra is cheaper; it is arguing that lower price plus speed can change which workflows are economically viable.
On MMMU-Pro, a visual-reasoning benchmark, OpenAI says GPT-6.1 Terra outperforms GPT-5.6 Terra at every reasoning setting and nearly matches GPT-6 Astra and GPT-6.1 Sol at much lower cost . On OSWorld 2.1, which tests how agents use desktop applications to complete multistep tasks, the company says Terra surpasses GPT-5.6 Terra’s best score at medium reasoning effort and at a fraction of the cost per task . Together, those claims define Terra as a model for visually grounded action: it is meant to understand images, screens and documents well enough to operate within real workflows.
Factuality and coding improvements
OpenAI also uses the launch to emphasize factuality. The company says it has invested heavily this year in reducing hallucinations and factual errors, and reports that an internal benchmark of error-inducing prompts showed GPT-6.1 Terra’s error rate at low reasoning effort falling from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra . At maximum effort, OpenAI says the error rate fell by 18 percentage points to 4.3%, nearly matching GPT-6 Astra and GPT-6.1 Sol .
For coding, OpenAI says GPT-6.1 Terra is a substantial improvement over GPT-5.6 Terra. On Terminal-Bench 4.0, a command-line coding benchmark, OpenAI says Terra solves 50.3% of tasks at maximum reasoning effort, more than double GPT-5.6 Terra’s published score of 21.5% . The company nevertheless keeps a clear hierarchy: for the most complex coding tasks, it says GPT-6.1 Sol and GPT-6 Astra remain better choices .
That distinction is important. Terra is not being sold as the absolute top model for every hard software-engineering problem. Instead, it appears designed for the many coding and technical operations tasks where the decisive factors are cost, speed, sufficient reliability and the ability to act over tools. On Terminal-Bench-Science 0.1, which tests how agents use code in scientific research tasks, OpenAI says Terra surpasses GPT-6.1 Sol at lower cost and comes within five points of Opus 5.5 at less than one-fifth the estimated cost .
Pricing: the most disruptive part
The pricing may be the most important commercial detail. OpenAI says GPT-6.1 Terra’s standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens . Its pricing page also lists gpt-6.1-terra in the broader API pricing structure, placing it inside the GPT-6 model family’s current cost ladder .
OpenAI says Terra’s standard API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol . That is a meaningful cut because agentic workloads can be token-intensive. A workflow that reads documents, navigates screens, uses tools, reasons over state and writes outputs may generate far more tokens than a simple chatbot exchange. Lower input, cached-input and output prices therefore have an outsized effect on whether developers can deploy agents at scale.
For businesses, the cached-input price is particularly relevant. Many enterprise workflows reuse large shared instructions, policies, schemas, tool descriptions and context blocks. If those tokens can be cached cheaply, the marginal cost of repeated tasks falls. Terra’s pricing suggests OpenAI is trying to make persistent or semi-persistent agent workflows less expensive, especially for high-volume operations such as invoice handling, catalog management, quality checks, technical documentation and form processing.
Availability in ChatGPT Work, Codex and the API
OpenAI says GPT-6.1 Terra is available starting today to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex . Developers can access it through the OpenAI API under the gpt-6.1-terra model ID . The model page confirms the API-facing use cases and notes that tool calling requires the Responses API .
This split is notable. Terra is being introduced directly into work-oriented environments rather than only into general chat. ChatGPT Work and Codex are exactly where users expect the model to manipulate documents, reason over software projects and assist in multi-step technical execution. That makes the deployment channel consistent with the model’s positioning: it is a workhorse for applied tasks.
Safety and alignment
The safety section of the launch is also significant. OpenAI says GPT-6.1 Terra meets the Critical threshold in cybersecurity under its Preparedness Framework and maintains the protections developed for GPT-6 Astra . The Deployment Safety Hub lists an October 26, 2026 addendum to the GPT-6 Astra system card for GPT-6.1 Terra, indicating that OpenAI is treating the release as part of the same safety-documentation lineage as its higher-tier GPT-6 models .
OpenAI also says Terra inherits alignment improvements from the GPT-6 series. According to the company’s evaluations, Terra is more likely than GPT-5.6 Terra to stay within scope and honestly disclose limitations or blockers . OpenAI also says Terra showed strong resistance in prompt-injection tests against malicious instructions designed to divert it from a user’s task .
Those details are especially relevant because Terra is built for computer use. A model that reads screens and acts inside applications has a different risk profile from a passive text generator. Prompt injection, over-delegation, unauthorized actions and misinterpretation of user intent become more consequential when the model can operate tools. OpenAI’s safety claims therefore address the right category of risk, even though outside auditors and real-world usage will ultimately determine how robust those protections are.
Why Terra matters
GPT-6.1 Terra matters because it reflects a broader shift in frontier AI: the contest is no longer only about the most capable flagship model. It is also about how much capability can be delivered per dollar, per second and per workflow. OpenAI’s announcement repeatedly links Terra’s benchmark results to lower task cost, faster completion and more efficient token use . That emphasis suggests the company sees the next phase of AI adoption as operational rather than experimental.
If OpenAI’s claims hold up in production, Terra could become a default choice for companies that want advanced reasoning and visual understanding but cannot justify using the most expensive model for every step. The higher-end models may still handle the hardest planning, architecture, research and debugging problems. Terra’s role is different: it aims to make everyday agentic work cheap and fast enough to repeat thousands or millions of times.
That is why the launch should be read as both a product release and a market move. GPT-6.1 Terra is OpenAI’s answer to a practical question: how do you make advanced AI useful not just in demos, but in the repetitive, document-heavy, screen-based workflows where businesses spend most of their time? OpenAI’s answer is a model that trades the glamour of the top tier for the economics of scale.