Technology

OpenAI's New AI Model Is Cheaper and Faster, but Its Maker Flags a Security Concern

Martin HollowayPublished 2w ago3 min readBased on 4 sources
Reading level
OpenAI's New AI Model Is Cheaper and Faster, but Its Maker Flags a Security Concern
Image by Pexels from Pixabay

OpenAI has released GPT-6.1 Terra, a fast and efficient AI model that can read images, operate a computer on a user's behalf, and write in natural language. The company says it is a significant step up from GPT-5.6 Terra for writing and computer-based work. OpenAI

OpenAI lists the standard API rates for GPT-6.1 Terra at 50% below those of GPT-5.6 Terra and GPT-6.1 Sol. An API is the connection developers use to build their own software on top of a model, and they pay for each use. The company also says GPT-6.1 Terra completes tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol.

OpenAI has also published its own test results. Benchmarks work a bit like a school exam: every model answers the same questions, so the scores can be compared.

On Agents' Last Exam, which tests AI systems that carry out multi-step tasks on their own, GPT-6.1 Terra scores 15 percentage points higher than GPT-5.6 Terra. It costs about one-third as much to run.

On MMMU-Pro, a test of reasoning over images and text, it beats GPT-5.6 Terra at every reasoning setting.

On Terminal-Bench 4.0, which tests tasks typed into a command line, it solves 50.3% of tasks when given its maximum reasoning effort. GPT-5.6 Terra solved 21.5%.

On OpenAI's own set of prompts designed to make models fail, the error rate fell from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra. This test used low reasoning effort.

OpenAI also reports that GPT-6.1 Terra meets the Critical threshold in cybersecurity under its Preparedness Framework. That is the company's own system for rating risk. Any team deciding where to use the model in security-sensitive work will need to take that rating into account.

Separately, OpenAI introduced GPT-6.1 Sol as an upgrade to GPT-6 Sol. The company says it nearly matches GPT-6 Astra on agentic coding, computer use and professional work. On September 29, 2026, OpenAI posted an availability update naming GPT-6.1 Sol as its latest model. OpenAI

The lineup has changed quickly. In an earlier preview of the GPT-5.6 models, OpenAI described Terra as a balanced model for everyday work. It listed Terra at $2.50 for input and $15 for output, where input is what you send the model and output is what it sends back. Luna was listed at $1 for input and $6 for output. The company said Sol, Terra and Luna would become generally available in the weeks after the preview. OpenAI GPT-6 Sol and GPT-6 Luna are available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users.

In my view, the numbers that matter most to engineers are the ones that can be compared fairly. The jump on Terminal-Bench, from 21.5% to 50.3%, is large enough to change how teams think about AI agents that run command-line tasks without supervision. A single score on a single test, though, tells less than the cost of each finished task, the speed under heavy use, and the mistakes that remain.

The broader context here is the price. A model that is both cheaper and quicker to finish a job changes what is affordable to keep running. Long jobs that were too expensive, such as multi-step computer sessions and reviews of large piles of documents, become easier to justify on the same budget. The Critical cybersecurity rating points the other way. It asks the same teams to add access controls, logging and monitoring before they widen how the model is used.

For now, this is a new model with aggressive pricing, strong reported gains on tests of agent and computer tasks, and a security rating that organisations will need to plan around. Whether these results hold up in everyday use is what the coming months will show.