A new Terra for the GPT-6 line
OpenAI has launched GPT-6.1 Terra, positioning it as a major upgrade to the Terra tier for ChatGPT Work, Codex and developers using the API . The model is described by OpenAI as a fast and efficient system designed for visual reasoning, precise computer use and clearer natural-language writing . In practical terms, this is not a broad replacement for every ChatGPT experience; OpenAI’s product notes say Terra is available in ChatGPT Work and Codex, and that it is not available in regular Chat .
That distinction matters. The release is aimed squarely at workflows where an AI agent needs to understand a screen, interpret a file, take action in software, or produce concise professional writing. OpenAI’s examples include drafting proposals, interpreting technical drawings, filling out forms and editing videos . In a separate product update, OpenAI says users can choose GPT-6.1 Terra from the model list in Advanced in the ChatGPT desktop app or in ChatGPT Work on the web, while Work also supports Terra on mobile .
The launch also fills a gap in OpenAI’s current model ladder. 9to5Mac notes that OpenAI’s main model classes now range from lower-cost Luna through Terra and Sol to the highest-end Astra, and that Terra had been absent when OpenAI released GPT-6 Sol and Luna last month . With GPT-6.1 Terra, OpenAI is reviving the middle option: more capable than the cheapest tier, less expensive than the premium models, and optimized for high-volume professional tasks.
Lower price, higher work capacity
The central business argument for GPT-6.1 Terra is cost efficiency. OpenAI says Terra advances the “cost-intelligence frontier” by offering a new balance of speed, capability and price . Its standard API prices are $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens . OpenAI says those standard API prices are 50% lower than GPT-5.6 Terra and GPT-6.1 Sol .
For ChatGPT Work and Codex subscribers, the point is not only cheaper API calls. OpenAI says the same usage allowance covers more tasks with GPT-6.1 Terra than with GPT-5.6 Terra, and substantially more than with GPT-6.1 Sol or GPT-6 Astra . That makes the release especially relevant for developers and teams who hit usage limits while running coding agents, browser automation, or document-heavy workflows.
This is why the model’s placement is important. Sol and Astra remain the premium choices for the most complex coding and agentic work, while Terra is meant to make frequent, repetitive and visually grounded work less costly. OpenAI gives the example of developers automating high-volume tasks such as reading invoices and entering details into accounting software . The value proposition is simple: not the absolute largest model, but a model that is good enough for many agentic tasks and cheap enough to run more often.
Visual reasoning and computer use are the headline capabilities
OpenAI’s announcement emphasizes visual understanding and action. GPT-6.1 Terra is presented as a model that can reason about images and screens, then act across browsers, desktops and apps . The company’s product notes describe possible uses such as matching supplier product photos to catalog entries, updating listings, checking campaign assets for cropping and text placement, and organizing edits .
OpenAI also reports benchmark improvements in computer-use tasks. On Agents’ Last Exam, a benchmark of professional tasks, OpenAI says GPT-6.1 Terra at medium reasoning effort scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost . OpenAI further claims that Terra performs on par with GPT-6 Astra on that benchmark while costing one-tenth as much per token . At maximum effort, OpenAI says Terra is more than three times as token-efficient as Claude Opus 5.5 and one-fourth its token price, producing a roughly nine-times lower cost per task .
OpenAI also says Terra is quicker, completing tasks more than three times faster than both GPT-6 Astra and GPT-6.1 Sol . 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 desktop-app use across multistep tasks, OpenAI says Terra at medium reasoning effort surpasses GPT-5.6 Terra’s best score at a fraction of the cost per task .
These are OpenAI’s own evaluations, so teams should still run their own tests before replacing existing production models. But the pattern of the claims is consistent: Terra is being pitched less as a conversational chatbot upgrade and more as an efficient action model for visual, desktop and workflow automation.
Factuality and coding: the quieter but important upgrades
The Terra release is also a factuality story. OpenAI says it has invested heavily this year in reducing hallucinations and factual errors, and the GPT-6.1 Terra numbers are presented as evidence of that push . On OpenAI’s internal benchmark of error-inducing prompts, the error rate at low reasoning effort reportedly fell 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 to 4.3%, nearly matching GPT-6 Astra and GPT-6.1 Sol .
For Codex users, the coding gains may be just as important. OpenAI says Terminal-Bench 4.0 measures how well coding agents solve command-line tasks, such as fixing a webpage whose layout shifts as it loads . At maximum reasoning effort, GPT-6.1 Terra reportedly solves 50.3% of Terminal-Bench 4.0 tasks, more than double GPT-5.6 Terra’s published score of 21.5% . OpenAI still cautions that GPT-6.1 Sol and GPT-6 Astra remain better choices for the most complex coding tasks .
On Terminal-Bench-Science 0.1, which tests code use in scientific research tasks, OpenAI says GPT-6.1 Terra at maximum reasoning surpasses GPT-6.1 Sol at lower cost and comes within five points of Claude Opus 5.5 at less than one-fifth the estimated cost . That makes Terra potentially attractive for research-adjacent automation, analysis pipelines and structured technical tasks, provided users validate outputs carefully.
Availability: ChatGPT Work, Codex and API
OpenAI’s API changelog records the Oct. 26 release of GPT-6.1 Terra for the Responses API and Chat Completions API . The model name for developers is gpt-6.1-terra . For consumer and business users, OpenAI says Terra is available in ChatGPT Work and Codex . The launch post lists availability for Plus, Pro, Enterprise and Edu users, while OpenAI’s ChatGPT Learn page also lists Business among the plans for ChatGPT Work and Codex .
There are also deployment caveats. OpenAI says Enterprise access depends on workspace availability and administrator controls . The company also notes that browser use and computer use each have setup and access requirements, and that GPT-6.1 Terra does not support UAE inference residency . These limitations mean the rollout may feel different depending on account type, workspace configuration and region.
9to5Mac summarized the consumer-facing significance succinctly: the new model is available through ChatGPT Work, Codex and the API, and it gives OpenAI a new medium-sized model that balances intelligence and affordability . For end users, the key test will be whether Terra makes Work and Codex feel less constrained without noticeably reducing quality.
Safety and preparedness disclosures
OpenAI published a system-card addendum alongside the release, and it is unusually important because Terra is not just a writing model: it is designed for agentic computer use . The addendum says GPT-6.1 Terra uses the same types of data and training as GPT-6 Astra and provides updated safety metrics for the new model . OpenAI says it is treating GPT-6.1 Terra as Critical in cybersecurity and High in biological and chemical capability under its Preparedness Framework .
Those labels do not mean the model is being released without safeguards. OpenAI says GPT-6.1 Terra uses the same safeguards stack as GPT-6.1 Sol and GPT-6 Astra, with probes updated for the model . The system-card addendum says Terra is intended to be used according to OpenAI’s usage policies, service terms and terms of use .
The addendum also provides useful context on risk. OpenAI says GPT-6.1 Terra performs on par with recent GPT-6 and GPT-6.1 releases in agentic safe-completion evaluations . In deployment simulations, OpenAI says Terra received approximately 42% fewer severity-level-3-or-higher flags than GPT-6 Sol and 64% fewer than GPT-5.6 Sol . On the other hand, OpenAI’s preparedness section says the company is treating Terra as Critical in cybersecurity, the same broad category as GPT-6.1 Sol and GPT-6 Astra .
That combination captures the release’s tension. Terra is meant to make agentic AI cheaper and more widely useful, but broader agentic access increases the importance of monitoring, prompt-injection resistance, administrative controls and user discipline.
Why this launch matters
GPT-6.1 Terra is not OpenAI’s biggest model, and that is the point. It is a middle-tier upgrade built for volume: enough intelligence for visual reasoning, coding assistance and computer-use tasks, but priced and tuned for more frequent deployment. If OpenAI’s claims hold up in real-world use, Terra could become the default choice for many everyday Codex and ChatGPT Work jobs where Astra or Sol are unnecessarily expensive.
The most important caveat is that users should not treat the launch as a blanket ChatGPT upgrade. OpenAI’s own product notes say the model is not available in Chat . This is a Work, Codex and API release, aimed at professional automation rather than casual conversation.
Still, it marks a meaningful step in OpenAI’s product strategy. GPT-6.1 Terra brings the GPT-6 generation into the middle of the lineup, lowers the price of agentic work, and gives teams a new option between low-cost volume and frontier capability. For developers, the model to test is gpt-6.1-terra; for ChatGPT users, the place to look is ChatGPT Work and Codex .