GPT-6.1 Terra
Reportedly the codename Terra for a next-generation model in OpenAI's GPT series, appearing in Microsoft evaluation documents. Its specific capabilities and specifications have not been officially confirmed.
Core Facts
Timeline (last 90 days)
The creator reminds that websites generated by Terra still need their copy, links and layout checked and modified before being used directly as commercial sites.
According to OpenAI's release notes, Plus, Pro, Enterprise and Edu users can access Terra in ChatGPT Work and Codex.
The creator said that after one quota reset, he switched to Luna 6's extra high tier to finish the editing, and Luna 6 at extra high performed better than Terra 6.1 at low reasoning effort at much lower cost, and he plans to stick with this combination going forward.
When the creator used GPT-6.1 Terra to make videos, it consumed usage credits far faster than expected.
In Ruben Soto's early comparison tests, Terra ran noticeably faster and at lower cost, but the results may still need to be run a few more times.
According to data published by OpenAI, on the Agents Last Exam benchmark, at medium reasoning effort, GPT-6.1 Terra scored 57.6% and GPT-5.6 Terra scored 42.6%, an improvement of 15 percentage points.
When releasing GPT-6.1 Terra, OpenAI emphasized computer use capabilities, that is, the model can observe screen content and interact with software, including clicking buttons, filling out forms, and advancing tasks across applications.
OpenAI's Ted Sanders stated in a Hacker News comment that GPT-6.1 Terra is likely better than Astra and Sol at technical communication, performs similarly on vision and computer use but is much faster, and is slightly weaker in most other respects
Simon Willison observed that the keywords OpenAI chose for GPT-6.1 Terra are visual reasoning and computer use, rather than general intelligence or reasoning depth
Ted Sanders said that for subscribers GPT-6.1 Terra is not a major release, and ordinary subscribers have no pressing reason to upgrade
26 more timeline events
All Facts (20)
OpenAI's official website published an article titled "Introducing GPT-6.1 Terra," unveiling a GPT-6.1 model named Terra.
65%UnverifiedThe creator said that after one quota reset, he switched to Luna 6's extra high tier to finish the editing, and Luna 6 at extra high performed better than Terra 6.1 at low reasoning effort at much lower cost, and he plans to stick with this combination going forward.
50%UnverifiedWhen the creator used GPT-6.1 Terra to make videos, it consumed usage credits far faster than expected.
50%UnverifiedIn Ruben Soto's early comparison tests, Terra ran noticeably faster and at lower cost, but the results may still need to be run a few more times.
50%UnverifiedAccording to data published by OpenAI, on the Agents Last Exam benchmark, at medium reasoning effort, GPT-6.1 Terra scored 57.6% and GPT-5.6 Terra scored 42.6%, an improvement of 15 percentage points.
50%UnverifiedWhen releasing GPT-6.1 Terra, OpenAI emphasized computer use capabilities, that is, the model can observe screen content and interact with software, including clicking buttons, filling out forms, and advancing tasks across applications.
50%UnverifiedAccording to OpenAI's release notes, Plus, Pro, Enterprise and Edu users can access Terra in ChatGPT Work and Codex.
50%UnverifiedThe creator reminds that websites generated by Terra still need their copy, links and layout checked and modified before being used directly as commercial sites.
50%UnverifiedSimon Willison observed that the keywords OpenAI chose for GPT-6.1 Terra are visual reasoning and computer use, rather than general intelligence or reasoning depth
50%UnverifiedTed Sanders said that for subscribers GPT-6.1 Terra is not a major release, and ordinary subscribers have no pressing reason to upgrade
50%UnverifiedSimon Willison compiled a comparison grid covering all reasoning effort levels across OpenAI's current frontier product line, including GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, and GPT-6 Luna
50%UnverifiedOpenAI's Ted Sanders stated in a Hacker News comment that GPT-6.1 Terra is likely better than Astra and Sol at technical communication, performs similarly on vision and computer use but is much faster, and is slightly weaker in most other respects
50%UnverifiedSimon Willison points out that GPT-6.1 Terra's positioning in OpenAI's product line is similar to that of the Sonnet model in Anthropic's model lineup
50%UnverifiedGPT-6.1 Terra achieves a vendor-reported score of 50.3% on Terminal-Bench 4.0.
50%UnverifiedGPT-6.1 Terra's output token unit price is five times that of input, so the cost of long-output tasks (such as code generation and long-form writing) is amplified.
50%UnverifiedThe prudent approach to tiered routing is to send simple requests to Terra, escalate complex or high-risk requests to Sol, and set up failure fallback at the routing layer.
50%UnverifiedUnder the medium-intensity setting of the Agents' Last Exam benchmark, GPT-6.1 Terra's score is on par with GPT-6 Astra.
50%UnverifiedUnder the medium-intensity setting of the Agents' Last Exam benchmark, GPT-6.1 Terra costs $0.29 per completed task, while GPT-6 Astra costs $4.82.
50%UnverifiedIt is currently impossible to confirm whether GPT-6.1 Terra comes with API updates, pricing adjustments or changes to the safety framework.
50%UnverifiedJudging from the version number "6.1," GPT-6.1 Terra looks more like an incremental update than a generational leap.
50%