DeepMind unveils Gemini 3.5 Flash Cyber for faster vulnerability scans

Google DeepMind has just rolled out Gemini 3.5 Flash Cyber, a lightweight model aimed at helping security teams find and fix bugs sooner. The company says it boosts code-path coverage and surfaced Chrome and Android flaws, but results await independent verification.

•
GoogleGeminiSecurity
google cover

TL;DR

  • Gemini 3.5 Flash Cyber: Specialized, lightweight model for spotting and patching vulnerabilities before exploitation
  • Access rollout: Limited pilot starting with governments and trusted partners
  • Speed and efficiency: Enables defense specialists to check significantly more code paths
  • Internal testing: Found complex vulnerabilities in @googlechrome and @Android that standard models missed
  • Chrome Production Commit Scanning Pipeline: 72% vs Gemini 3.5 Flash 55% and Opus 4.6 54%, per DeepMind material

Google DeepMind on Thursday introduced “Gemini 3.5 Flash Cyber,” a “specialized, lightweight model” the company claims is designed to help security teams “spot and patch vulnerabilities before they can be exploited.” DeepMind also said access will start with a limited pilot for governments and trusted partners.

In a follow-up post, the company stated that the model’s speed and efficiency let defense specialists check “significantly more code paths.” It further claimed that tests on its own codebases, including @googlechrome and @Android, found complex vulnerabilities that “standard models missed.”

A companion graphic titled “Gemini 3.5 Flash Cyber” and subtitled “Chrome Production Commit Scanning Pipeline” places the model at 72%, ahead of Gemini 3.5 Flash at 55% and Opus 4.6 at 54%. Those figures come from Google DeepMind’s own material and should be treated with caution pending independent verification.

Source: Google DeepMind on X

Continue the conversation on Slack

Did this article spark your interest? Join our community of experts and enthusiasts to dive deeper, ask questions, and share your ideas.

Join our community

Related Articles

Google launches Gemini 3.5 Flash with 4x faster coding

Google launches Gemini 3.5 Flash with 4x faster coding

Google has just rolled out Gemini 3.5 Flash, touting "frontier-level" agentic and coding performance at 4x the speed and often under half the cost. It’s available now in the Gemini app, Search AI Mode, and developer tools, with mixed early reactions.

2 shared tags
Gemini CLI adds subagents for faster parallel AI coding

Gemini CLI adds subagents for faster parallel AI coding

Google has just rolled out subagents in Gemini CLI, letting developers delegate frontend, tests, and docs to specialized helpers running in parallel. Each subagent gets its own tools and context, while the main session stays focused and coordinated.

2 shared tags
claude cover

Anthropic brings Claude Mythos 5 to Claude Security beta

Anthropic has just rolled out Claude Mythos 5 in Claude Security, now in public beta for all Claude Enterprise customers. The service scans GitHub repos for vulnerabilities and suggests fixes, but Mythos 5 isn’t exposed for direct use. Pricing is standard token usage.

1 shared tag