DeepSeek V4.1 Flash 1017 tested: agentic coding jumps to flagship-tier, but turns conservative
karminski3 · x · 2026-10-20
karminski3 shares test results for deepseek-v4.1-flash-1017: on his KCORES Vector DB Bench (75 models, each building a high-performance vector DB from scratch within 50 tool calls, ranked by QPS at recall ≥95%), the new version's backend agentic coding jumps from roughly Grok-4.6 level to flagship-tier, scoring 9,358 QPS at 97.71% recall (#7). He questions whether v4.1-pro is still necessary.
Agent behavior changed significantly: the previous version was "proactive," the new one "conservative," requiring a well-built harness; budget control seems worse, but improves when the budget is emphasized each session. A detailed video is expected this week.
Related event: DeepSeek V4.1 Flash Tested: Agentic Coding Jumps to Domestic Flagship Level(2 posts)→
More from coding & agent
- REA gains 214K stars in a week, tops GitHub's fastest-growing AI repos — sven_ai · 2026-10-20
- Dev goes 'graph pilled': graphs will anchor every company's world model — _AustinCalvert_ · 2026-10-20
- RuneBench: Benchmarking AI Coding Agents by Playing RuneScape — charliermarsh · 2026-10-20
- A 5-step loop for debugging AI agents in production: trace everything, watch for silent failures — Tracify · 2026-10-20
- New Claude Cookbook: Auto-Review Agent Session Transcripts for Security Issues — brada · 2026-10-20
- Conversational Video Editing: Driving CapCut Web with Codex for Marketing Videos — Div_pradeep · 2026-10-20