LLMs finally write more secure code: Veracode benchmark average jumps 56% to 62%
WeldPond · x · 2026-10-30
Veracode's GenAI code security benchmark shows its first average improvement: the security pass rate jumped from 56% to 62%.
- The new high-water mark is 73%, shared by Claude Opus 5.5 and GPT-6-Sol; Claude Sonnet 5.5 follows at 72%, Grok 4.7 at 69%;
- Ten models now clear 60%, up from four, including Qwen 3.8-Max, GLM 5.3 and Gemini 3.8 Flash;
- Cross-site scripting (CWE-80), a long-standing weak spot, is finally climbing.
Author's takeaway: models are starting to write more secure code — getting it right the first time saves fixing and verifying security bugs.
Related event: AI-Generated Code Security Pass Rate Rises for First Time: Veracode(3 posts)→
More from coding & agent
- Skyworks cuts root cause analysis from 20 days to 10 with LangGraph agents — LangChain · 2026-10-30
- 73-entry list sorts agent infra: which MCP tools an agent can provision itself vs needs a human — Total-Reasonable · 2026-10-30
- YC-backed Turnstone turns your existing AI subscriptions into an agent team that auto-switches when limits run out — ycombinator · 2026-10-30
- The simple self-testing loop behind every RL data company, per practitioners — teodorio · 2026-10-30
- 7 AI models all faked completion reviewing the same 111k-line codebase — a postmortem — rseroter · 2026-10-30
- Supabase mistook thousands of new databases per minute for a DDoS — it was Lovable integrating silently — alifcoder · 2026-10-30