Google Launches Lower-Cost AI Security Model Gemini 3.5 Flash Cyber
Decision Brief
Google released Gemini 3.5 Flash Cyber, a specialized security model based on Gemini 3.5 Flash, positioned as a cost-effective alternative to Anthropic's high-cost Mythos. It is initially available to government and trusted partners via CodeMender, Google's secure coding agent, which leverages fast, low-cost multiple invocations to scan more code paths and find vulnerabilities. In CyberGym AI security benchmarks, 3.5 Flash Cyber matched larger models with up to five calls; it identified 55 unique issues in the V8 JavaScript engine, surpassing 3.5 Flash's 47 and Opus 4.6's 36, with 10 undiscovered by others. Google also released Gemini 3.6 Flash (improved coding and multimodal) and 3.5 Flash-Lite (most cost-effective in the series). For security teams, 3.5 Flash Cyber offers lower costs, especially for those limited by Mythos's pricing. The multi-invocation design enables deeper code audits to uncover hidden bugs, but initial government-only access limits availability for developers.
Sources
- The Verge:AI
Consumer AI products, platform shifts, and policy/culture impact.
- The Verge:AI
留言
登入后即可留言,和其他 builder 交换实测心得。
还没有留言,抢头香。