GPT-6.2 tops browser-use benchmark: 82% success, 35% cheaper, 2x faster
gdb · x · 2026-11-18
- browseruse leaderboard: GPT-6.2 Sol is now the top model for browser use at 82% success rate, up from 81% for GPT-6.1 Sol.
- Efficiency gains: 35% cheaper, 2x faster (12 min vs 22 min per task), and 36% fewer steps.
- gdb calls GPT-6.2 "incredible at browser use."
Related event: GPT-6.2 tops browser-use leaderboard at 82% success(2 posts)→
More from coding & agent
- Perplexity launches Context API: clean markdown from URLs at $1 per 1,000 fetches — perplexity_ai · 2026-11-19
- Replacing a $150K/yr media buyer's weekly ad review with 10 Claude skills across 4 ad platforms — aryanXmahajan · 2026-11-19
- Dev open-sources SimplePost: let your ChatGPT/Claude agent schedule posts to 10 networks — haltakov · 2026-11-19
- Vibe-coded in 4 days: a 150-million-star galaxy in one 299 kB HTML file — BigTutor6739 · 2026-11-18
- Stateless Streamable HTTP MCP server on Cloudflare Workers (19ms) to shield FX agents from news whipsaws — No_Comedian_6913 · 2026-11-18
- After 8,000+ Claude sessions, he rebuilt Notion, Maps and Finder into one open-source agent OS — Conchylicultor · 2026-11-18