DeepSeek V4.1 Flash 1017 finds 54 vulns for $7.53, topping open-weight leaderboard
zizhpan · x · 2026-10-17
BugBunny.ai tested DeepSeek's updated V4.1-Flash-1017 on the VulnPR-100 benchmark (100 PRs with known vulnerabilities, one hour per PR): it found 54 vulnerabilities for just $7.53 ($0.14 per vuln), making it the best open-weight model on the leaderboard.
- Matches Grok 4.6's 54 findings at roughly 16× lower cost; only 3 findings behind Gemini 4 Argon (57) at 22× less
- Previous V4.1 Flash found 44 for $12.79 — the update adds 10 more findings while cutting cost 41%
- Full leaderboard in charts: Grok 4.7 leads at 62/100 ($336.67), Claude Opus 5 at 48/100 ($447.90)
More from coding & agent
- Ask Reddit: How do you evaluate production AI agents — mock tools or real calls? — Mountain_Edge1061 · 2026-10-18
- Studio H3: open-source AI film studio with full director workflow — Exciting-Income-5840 · 2026-10-18
- Agent teams with per-agent sandboxes, MCPs, skills and email inboxes for task delivery — pramodk73 · 2026-10-18
- How do you test a RAG support bot for wrong answers before it ships? — srinath_y · 2026-10-18
- Snorkel AI: 4B model beats 235B on financial analysis for under $500 with RL and verified data — ajratner · 2026-10-18
- User praises Grok's X-data access but switches back to Claude Code for heavy work — jacob_posel · 2026-10-18