gpt-6.1-terra aces screen reading and JSON in hands-on test

GPT-6.1-terra is the speed-and-cost option inside the GPT-6.1 family, with Luna aimed at lightweight bulk processing, Sol at complex work, and Astra at the top. That positioning matters because the author's tests are deliberately scoped to tasks where speed and cost matter more than deep reasoning: reading a screen, planning a short procedure, extracting JSON, and writing a brief intro.

The screen-reading test used a dark orbit-animation scene with a central glowing sphere, six labeled points, FPS and speed readouts, and buttons and a slider. The model captured all of those details correctly, which the author treats as evidence that pulling on-screen text and positions is practical, while still recommending spot checks of important items before production use.

On planning, the model returned three steps for enabling a data-sharing switch, and in an overwrite-confirmation scenario it chose 'save as' with report_v2.pdf rather than the default overwrite, adding a confirmation step. The author notes that the model does not click by itself and that risky actions should be human-confirmed on the execution side. The JSON and drafting tests round out the picture: parseable output with every requested field, and a readable short intro that still needs human review for tone and unsupported claims.