Overview
- Google released Gemini 3.8 Flash and a Cyber variant on Wednesday, September 2, 2026, making the standard Flash model broadly available and restricting the Cyber version to vetted defenders via the Fairwind Program.
- The company kept introductory per‑token pricing through December 31, 2026 at $0.75 per million input tokens and $3.75 per million output tokens while warning that higher effort settings make the model 'work harder' and may use more tokens per task.
- Google published benchmark gains for 3.8 Flash on long‑horizon coding (DeepSWE v1.1), multi‑step reasoning (HLE‑Verified 54.9%), and other professional tests and said 3.8 Flash Cyber matches or nears frontier patching scores on CWE‑Bench and CyberGym.
- Independent third‑party replications of Google’s head‑to‑head claims remain limited, and Google acknowledges tradeoffs: developers can lower effort settings or continue using 3.7 Flash for efficiency if token usage is a priority.
- This is Google’s third Flash release in about six weeks, reflecting a tactical shift to iterate smaller, lower‑latency models for developer and agent workloads while larger Gemini Pro models face delays and more costly compute needs.