OpenAI Launches GPT-6.1 Terra: Computer Use at Half the Price

October 28, 2026 · Cortex Solutions

OpenAI released GPT-6.1 Terra on October 26: a computer-use model at 50% lower API prices. What it can do, the caveats, and what it means for Saudi firms.

On Monday, October 26, 2026, OpenAI released GPT-6.1 Terra, the middle model in its lineup, focused on what the company calls "computer use": the model sees the screen, understands what is on it, then clicks, types and moves between apps the way an employee would. According to coverage by BleepingComputer (Mayank Parmar, October 26, 2026), Decrypt (Jose Antonio Lanz, October 26, 2026) and CXO Digitalpulse (October 27, 2026), the headline of the launch is price: the model costs 50% less than its predecessor.

What is GPT-6.1 Terra?

OpenAI describes the model, as quoted by BleepingComputer from its blog, as "a fast, efficient model that combines detailed visual reasoning and precise computer use." It is aimed at everyday office work: filling out forms, editing videos, interpreting technical drawings and automating repetitive desktop tasks, as well as coding and writing.

Decrypt notes that Terra first appeared on June 26 as part of the GPT-5.6 family in a limited preview, then opened to the public on July 9. OpenAI describes the new release as "a significant step up from GPT-5.6 Terra."

Pricing and availability

According to all three sources, the model is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work (ChatGPT's agent mode) and in the Codex coding tool, and developers can use it through the API as gpt-6.1-terra. Decrypt notes that free users get the lower-tier GPT-5.6 Luna.

Standard API pricing, according to CXO Digitalpulse, is:

• $1 per million input tokens.
• $0.05 per million cached input tokens.
• $5 per million output tokens.

OpenAI says these rates are 50% below the standard pricing of GPT-5.6 Terra and GPT-6.1 Sol. CXO Digitalpulse also reports a context window of 1.05 million tokens and a knowledge cutoff of June 1, 2026.

Performance, according to OpenAI

The sources reported figures published by OpenAI itself, the most important being:

• On Agents' Last Exam at medium reasoning effort, the model scored 15 percentage points higher than GPT-5.6 Terra at about one-third of the cost, and the company says it performs on par with GPT-6 Astra at one-tenth the price per token.
• It completes tasks more than three times faster on average than GPT-6 Astra and GPT-6.1 Sol.
• On Terminal-Bench 4.0, a command-line coding benchmark, at maximum reasoning effort it solved 50.3% of tasks versus 21.5% for GPT-5.6 Terra.
• On an internal test of prompts designed to mislead models, the factual error rate at low reasoning effort fell from 53.6% to 16.7%, and at maximum effort it fell 18 points to 4.3%.

Even so, OpenAI still recommends GPT-6.1 Sol or GPT-6 Astra for the most complex coding tasks and in-depth research.

Caveats worth noting

Decrypt points out that all performance figures come from OpenAI and that the error test is internal and not independently verified. CXO Digitalpulse likewise says the results have not yet been reproduced by an outside party, and that real-world performance depends on latency, tool reliability, security controls and application-specific evaluation.

On security, CXO Digitalpulse reports that OpenAI classifies the model at the "Critical" cybersecurity threshold under its Preparedness Framework, explaining that this reflects advanced capability rather than a disclosed incident, and that it puts more weight on access controls, monitoring and deployment safeguards.

What this means for Saudi businesses

The clearest message from this launch is that the cost of AI agents that work directly on the screen is falling fast. That opens the door for Saudi companies, small or mid-sized, to automate repetitive office work that was expensive only a few months ago. We recommend the following:

• Start with one clear, repetitive task, such as entering data from documents into an internal system or filling in standard forms, and measure the result before scaling.
• Do not rely on vendor figures alone; test the model on your own data and tasks, and check its accuracy in Arabic specifically.
• Give the agent the fewest permissions possible, and keep human approval before any financial action or data deletion, especially given the model's Critical cybersecurity classification.
• Check where your data is processed, and comply with the Kingdom's Personal Data Protection Law before sending customer data to any external service.
• Compare the real cost: a lower price per token does not always mean a lower bill if a task needs many attempts.

At Cortex Solutions we help companies design AI agents and automate their operations with clear controls and human approval at sensitive points. Learn about our service: https://cortex.sa/services/ai-automation