DeepSeek V4.1 Flash 1017 matches Opus 5.5 on long-horizon browser use at 19x lower cost
zizhpan · x · 2026-10-18
The Browser Use Benchmark v2.1 (180 shared tasks) shows DeepSeek V4.1 Flash 1017 scoring 75/100 on long-horizon browser use, matching Opus 5.5 and nearing Astra level — at $0.22/task vs $2.35/task, roughly 19x cheaper. The accompanying Pareto chart adds more detail: the low setting hits 74.0 for just $0.12, while GPT-6.1 Sol (medium) remains SOTA at 79.6 for $1.30. DeepSeek's new model effectively redraws the cost-performance frontier for browser agents.
More from coding & agent
- An Agent Can Remember Perfectly and Still Be Wrong: What Evidence Proves 'Done' After a Context Reset? — jylusdev · 2026-10-18
- Grok Bot vs Claude Code: When Compute Is Elastic, Platform Capabilities Decide — tekbog · 2026-10-18
- A prompt to make your agent audit under-used features across your MCP-connected tools — thejasminejade · 2026-10-18
- Writer has Claude redesign her Obsidian plugin UI with zero backend changes — evielync · 2026-10-18
- AI agents hunting a $465 train refund invented excuses, leaked medical data, and spammed an MP — The Decoder · 2026-10-18
- Open-source Grok Bot alternative cuts coding agent costs from $110/day to $4 — Arindam_1729 · 2026-10-18