GPT-6.1 Terra makes computer-use agents cheap enough for everyday automation

What GPT-6.1 Terra is

GPT-6.1 Terra is OpenAI's fast, efficient model for computer use and visual reasoning. It reads screens, interprets technical drawings, fills out forms and edits video, and OpenAI says it writes in a clearer, more natural voice than GPT-5.6 Terra, the model it succeeds. It is available now in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and in the OpenAI API as gpt-6.1-terra.

GPT-6.1 Terra API pricing

The standard API price is $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens. That is 50 percent lower than GPT-5.6 Terra and GPT-6.1 Sol.

The comparison that matters is cost per finished task. On Agents' Last Exam, a benchmark of professional computer work, Terra at medium effort performs on par with GPT-6 Astra, OpenAI's frontier model, at one tenth of Astra's price per token. Against Claude Opus 5.5 it is over three times as token efficient and a quarter of the price per token, which works out to about nine times lower cost per task. It also finishes tasks more than three times faster than Astra or Sol.

How capable GPT-6.1 Terra is

• On Agents' Last Exam at medium effort, it scores 15 points higher than GPT-5.6 Terra at about one third of the cost.
• On OSWorld 2.1, which tests agents operating real desktop apps, it beats GPT-5.6 Terra's best score at a fraction of the cost per task.
• On MMMU-Pro visual reasoning, it beats GPT-5.6 Terra at every reasoning setting and nearly matches Astra and Sol at much lower cost.
• On Terminal-Bench 4.0 for coding agents, it solves 50.3 percent of tasks at maximum effort, more than double GPT-5.6 Terra's 21.5 percent. For the most complex coding tasks, GPT-6.1 Sol and GPT-6 Astra remain the better choice.
• On OpenAI's internal factuality benchmark, the error rate fell from 53.6 percent with GPT-5.6 Terra to 16.7 percent at low effort, and to 4.3 percent at maximum effort, nearly matching the frontier models.
• On safety, it meets the Critical threshold in cybersecurity under OpenAI's Preparedness Framework and showed strong prompt-injection resistance in OpenAI's tests.

What GPT-6.1 Terra means for SaaS teams

The jobs that never cleared the cost bar now do. OpenAI's own example is reading incoming invoices and entering their details into accounting software, and it is the right shape: high volume, visual, repetitive, boring. Support-form triage, back-office data entry, claims processing and document review all fit the same pattern. Work that needed a frontier model's budget now runs at a tenth of the price, which turns a nice demo into a line item that pays for itself.

Two practical notes. First, cached input at $0.05 per million tokens means products that resend the same context, like a fixed tool list or a long system prompt, keep most of the discount automatically. Second, Terra is not the model for everything: for the hardest coding work, OpenAI itself points you to GPT-6.1 Sol and GPT-6 Astra. Route the bulk traffic to Terra and escalate the hard cases.

That routing pattern is how I build agent features into products in practice. My Agent Skills case study covers the reusable-instructions side of the same work.

FAQ

Sources

1. 01Announcement, benchmarks and pricing: Introducing GPT-6.1 Terra - OpenAI (opens in a new tab)

Elsewhere on this site

The selected work catalogue lists the SaaS products and dashboards behind notes like this one.