GPT-6.1 Terra Nearly Matches Sol on Browser Tasks

Terra’s score makes it the more compelling value choice for browser agents, even if Sol retains a narrow performance lead.

• –A 1.2-point gap is effectively parity on a noisy, long-horizon benchmark.
• –The 35% cost advantage could materially improve agent economics at scale.
• –Browser Use’s benchmark evaluates real websites with weighted task rubrics, making the result more meaningful than simple navigation tests.
• –Developers should default to Terra for routine browsing, form filling, and research workflows, reserving Sol for harder tasks.
• –The comparison remains setup-sensitive: reasoning effort, harness, browser environment, and judge configuration can shift the ranking.