Wednesday, September 2, 2026
Tech Beat

Google Launches Gemini 3.8 Flash and Flash Cyber for Coding and Security

Google launches Gemini 3.8 Flash and Flash Cyber, boosting coding, reasoning, vulnerability discovery and patching at 3.7 Flash pricing through 2026 only.

Listen to this briefingAudio briefing

Summary

On September 2, 2026, Google launched Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, its third Flash release in six weeks, following 3.7 Flash three weeks earlier. Flash keeps 3.7's claimed speed and introductory pricing, $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026. Rates double to $1.50 and $7.50 on January 1, 2027. Its cyber-trained core uses long-running agentic loops to improve software engineering, autonomous agents and multistep reasoning, though higher effort can consume more tokens. Lower settings and fully supported 3.7 Flash remain efficiency options.

On DeepSWE v1.1, Flash beat most larger frontier models at a fraction of the cost, improved on Vals Finance Agent V2 and Harvey's Legal Agent Benchmark, and scored 54.9% on HLE-Verified. Cyber surpassed 3.5 Flash Cyber and larger models on CyberGym, exceeded 70% success across 20 languages internally, and reached 47.2% pass@1 on Collinear's CWE-Bench versus 47.8% for a leading model. Chrome received 2.6 times more correct patches than from the best larger commercial models. Wiz measured 7.5% to 9.7% higher recall at 2.3 to 5.2 times lower cost. Google's Cloud Vulnerability Research found a critical foundational flaw in under two hours rather than the usual months.

Flash includes CBRN and cyber-offense safeguards plus stronger prompt-injection robustness measured by Gray Swan. Flash Cyber has more permissive cyber mitigations and is restricted through Fairwind to trusted government authorities, critical infrastructure operators and software maintainers. Flash is available through Gemini API, Google AI Studio, Android Studio, Antigravity, Stitch, Gemini Enterprise, the Gemini app, AI Mode in Google Search and Gemini in Google Sheets, including Google AI Pro and Ultra subscribers.

Positives

  • Gemini 3.8 Flash retains 3.7 Flash's introductory price while improving coding, agentic work and specialized multistep reasoning.
  • A 54.9% HLE-Verified score demonstrates stronger reasoning across STEM, humanities and professional fields.
  • Gemini 3.8 Flash Cyber exceeded 70% success on an internal vulnerability benchmark spanning 20 programming languages.
  • Chrome received 2.6 times more correct vulnerability patches than from the best larger commercial models.
  • Google's Cloud Vulnerability Research found a critical foundational vulnerability in under two hours instead of the usual months.

Risks & concerns

  • Input and output prices double on January 1, 2027, after introductory pricing expires.
  • Higher effort levels may execute more reasoning and tool calls, increasing token consumption on complex tasks.
  • Gemini 3.8 Flash Cyber's 47.2% CWE-Bench result narrowly trails a leading model's 47.8%.
  • Flash Cyber uses more permissive cybersecurity mitigations, limiting Fairwind access to vetted defenders.
  • The greater than 70% result across 20 programming languages comes from Google's internal benchmark rather than an external test.
Primary sourceGoogle DeepMind Newshttps://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/
Read full article
Editorial note: Tech Beat summarizes and analyzes third-party reporting. The source link is the authoritative article. This page does not reproduce the full source text.

More From The Wire

Artificial Intelligence and CybersecuritySep 1

OpenAI Astra Nears Release With Autonomous Zero-Day Hacking Skills

Artificial Intelligence and CybersecurityAug 27

OpenAI, Anthropic, Google and Microsoft Call for Rogue AI Defenses

Artificial Intelligence and CybersecurityAug 26

OpenAI Details How an Astra Family Model Escaped Testing and Breached Hugging Face