Simon is a Computer Science BSc graduate who has been writing about technology since 2014, and using Windows machines since 3.1. After working for an indie game studio and acting as the family's go-to technician for all computer issues, he found his passion for writing and decided to use his skill set to write about all things tech.
Since beginning his writing career, he has written for many different publications such as WorldStart, Listverse, and MakeTechEasier. However, after finding his home at MakeUseOf in February 2019, he would eventually move on to its sister site, XDA, to bring the latest and greatest in Windows, Linux, and DIY electronics.
Sign in to your XDA account
Summary
- GPT-6.1 Terra excels at multi-step computer tasks (Agent's Last Exam) and outperforms OpenAI models.
- It's close to Claude Opus 5.5 performance but far cheaper—$1.01 vs $8.82.
- Now available to Plus, Pro, Enterprise, Edu; 1M input tokens $1, 1M output tokens $5.
It seems a big pattern with GPT-6 is offering better performance for cheaper. Back when Anthropic revealed Claude Opus 5.5, OpenAI returned fire with GPT-6 Sol and Luna, which OpenAI claimed were both better yet more affordable than GPT-5.6. Now, OpenAI has pulled back the curtain and revealed GPT-6.1 Terra, which offers better computer use for less.
GPT-6.1 Terra is reportedly better than other OpenAI models at computer use
It scored pretty well in OpenAI's tests
On the OpenAI website, the company announced the release of GPT-6.1 Terra. OpenAI says it put 6.1 Terra through the paces with something called the "Agent's Last Exam," which is when an agent is tested to see if it can perform complex, multi-step computer tasks accurately.
You can see the results in the chart above. Each dot in a model's 'constellation' represents effort level, with the weakest on the left and the most powerful on the right. GPT-6.1 Terra on max effort—the dot on the far right of its line—overshadows all of OpenAI's models pretty effortlessly. It's also only 3.2 percentage points behind Claude Opus 5.5, while only costing $1.01 versus Claude's $8.82 bill.
Still, benchmarks are only a small window into how an LLM performs, and the real proof of adequacy will happen when people begin running it for their daily tasks. If you want to do that, GPT-6.1 Terra is now available to all Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, and you can get 1 million input tokens for $1 and 1 million output tokens for $5.