DeepSeek ships V4.1-Flash-1017, hits 77% on DeepSWE and opens Search API

dl_weekly · x · 2026-10-25

DeepSeek released DeepSeek-V4.1-Flash-1017 five weeks after the original, a minor update that cuts task execution time and output length. It scores 77% on DeepSWE v1.1 (up from 74.2%), edging past Claude Opus 5's 74%, though Anthropic still leads 7 of 11 published benchmarks. The update is tuned for agents including DeepSeek Harness, Tencent's CodeBuddy and WorkBuddy. DeepSeek also opened a Search API returning up to 50 ranked web results per query, plus a Fetch API converting pages to Markdown, at 75 cents per 1,000 calls.

Related event: DeepSeek V4.1 Flash 1017 tops open-weight cyber leaderboard, self-reported DeepSWE 77% beats Claude Opus 5(6 posts)→

Original post →

More from Models

• Meta Muse avatar sharing surfaces: four avatar states, social sharing, no release date — testingcatalog · 2026-10-25
• Meta Muse to add avatar sharing, hinting at an agent avatar store — testingcatalog · 2026-10-25
• Skeptic doubts rumored OpenAI model, citing inability to serve Astra reliably — CtrlAltDwayne · 2026-10-25
• AsciiBench: blind voting arena tests which AI draws better ASCII art across 13 models — JasonBotterill · 2026-10-25
• Satirical ARC-AGI-3 leaderboard shows GPT-6 near 90% score at $10K+ per task — haider1 · 2026-10-25
• Rumor: Anthropic to drop Opus 5.5 two weeks after calling for frontier pacing — StewartalsopIII · 2026-10-25

Models channel →