Kimi K3.1 Pro tops Artificial Analysis Cyber Index at 64, up 23 points over K3
ArtificialAnlys · x · 2026-10-24
Artificial Analysis updated its Cyber Index, a composite benchmark of model capability on cyber defense (finding, reproducing, and patching vulnerabilities in real repositories without breaking functionality).
- Kimi K3.1 Pro tops the index at 64, up 23 points from Kimi K3's 41; DeepSeek V4.1 Flash 1017 (Max) follows at 57, with Grok 4.7 (Xhigh) at 56.
- The gain is mainly driven by CyberGym-E2E-AA, jumping from 58% to 95% (+37 points) — the highest score of any model tested.
- The index combines three evaluations (CWE-Bench-AA, DeepsecBench-AA, CyberGym-E2E-AA) and is developed with the Cyber Index Alliance, whose partners include Collinear AI, IBM, NVIDIA, and Vercel.
Related event: Kimi K3.1 Pro tops open models on Artificial Analysis Cyber Index(2 posts)→
More from coding & agent
- Factory tested Reflection's Beam early via Together AI on real coding and agentic workflows — togethercompute · 2026-10-24
- Seroter's Reading List: CLIs Save Tokens, Agent Harness Explained, When to Use Diffusion LLMs — rseroter · 2026-10-24
- FrontierCS 2 benchmark makes agents beat original methods from code with the paper's contribution deleted — berkeley_ai · 2026-10-24
- Dev's workflow: pairing Grok bot with open-source Hermes Agent side by side — alexcovo_eth · 2026-10-24
- Anthropic's Claude Duet: a proactive agent that keeps working after you close the app — edwinarbus · 2026-10-24
- Workato's Google Workspace MCP uses one tool, 1,000+ AI-reviewed self-updating actions — adamse · 2026-10-24