Skip to main content
Latest Update
2026-11-18 07:57 UTC

GPT-6.2 Sol tops browser-use benchmark with faster, cheaper runs

GPT-6.2 Sol tops browser-use benchmark with faster, cheaper runs

According to Greg Brockman citing Browser Use on X, GPT-6.2 Sol achieved 82% on browser-use tasks, is 35% cheaper, 2x faster at 12 minutes, and needs 36% fewer steps than GPT-6.1 Sol.

Source

Analysis

According to Greg Brockman on November 18, 2026, GPT-6.2 Sol tops browser use benchmarks at 82 percent success, up from 81 percent for GPT-6.1 Sol, while cutting costs by 35 percent, halving runtime to 12 minutes from 22 minutes, and reducing steps by 36 percent as measured by Browser Use.

This advance in agentic browser capabilities directly improves developer and professional productivity by automating web-based research, data entry, and navigation workflows with fewer interventions. It also signals broader ecosystem shifts toward more efficient multimodal agents that integrate seamlessly into daily tools, reshaping how teams handle repetitive online tasks at scale.

Greg Brockman

@gdb

President & Co-Founder of OpenAI