Browser Use Puts Opus 5, Sol Neck-and-Neck

The frontier model race is increasingly task-dependent rather than winner-take-all.

• –GPT-5.6 Sol leads on some browsing and coding evaluations, while Opus 5 is highly competitive on computer-use and reasoning tasks.
• –Small differences in model effort, tool access, prompts, and evaluation harnesses can materially change the rankings.
• –Developers should benchmark models on their own workflows instead of relying on a single leaderboard.
• –Browser Use’s model-agnostic approach makes it useful for comparing frontier models under the same browser-agent setup.