Google on September 2 released Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, its third Flash-class model launch in six weeks2,5,13. The general-purpose Flash targets agentic coding and multi-step reasoning, while Flash Cyber is a restricted cybersecurity variant available only to vetted defenders through a new access program called Fairwind3.
Gemini 3.8 Flash arrives three weeks after Gemini 3.7 Flash10. Both 3.8 variants share the same foundational model but diverge in post-training: Flash is tuned for long-horizon coding and autonomous agents, while Flash Cyber is optimized for vulnerability detection and automated patching6. Google CEO Sundar Pichai said 3.8 Flash delivers "significant leaps" from 3.7 Flash across software engineering, agentic tasks, and multi-step reasoning.
On benchmarks, Gemini 3.8 Flash scored 73.7% on DeepSWE v1.1, outperforming most larger frontier models on long-horizon software engineering tasks. It achieved 54.9% on HLE-Verified and outperformed 3.7 Flash and other frontier models on Vals Finance Agent V2 and Harvey's Legal Agent Benchmark9. Google says Flash 3.8 often outperforms GPT-5.6 Sol, Claude Sonnet 5, and Claude Opus 5 on complex engineering tasks. In Arena.ai rankings, 3.8 Flash debuted at No. 7 in Text Arena, ahead of Claude Opus 5, and landed at No. 14 in Agent Arena, above DeepSeek-V4-Pro.
The model "works harder" than its predecessor by performing more reasoning steps and calling tools iteratively, which Google warns may consume more tokens at higher effort levels14. Developers who want to minimize token usage can remain on Gemini 3.7 Flash, which stays fully supported.
Gemini 3.8 Flash Cyber scored 86.2% on the CyberGym cybersecurity benchmark and 47.2% pass@1 on CWE-Bench11. On an internal Google benchmark spanning 20 programming languages, Flash Cyber achieved a vulnerability discovery success rate exceeding 70%. Google's Chrome Security team found that Flash Cyber produced 2.6 times more correct patches for Chrome vulnerabilities than the largest commercial models tested. Wiz reported 7.5–9.7% higher recall on its internal penetration-testing benchmark at 2.3–5.2 times lower cost than leading frontier models. Google's Cloud Vulnerability Research team used Flash Cyber to find a critical foundational vulnerability in less than two hours.
Flash Cyber is restricted to roughly 650 trusted partners through the Fairwind Program, including Accenture, CrowdStrike, the Center for Internet Security, Datadog, Palo Alto Networks, Snowflake, and Wiz. The program prioritizes government authorities, critical-infrastructure operators, and software maintainers15. Flash Cyber carries a more permissive set of cybersecurity mitigations than the standard Flash, while the standard model ships with safeguards against misuse in CBRN and cyber-offense domains.
Pricing for Gemini 3.8 Flash matches the 3.7 Flash introductory rate: $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026, rising to $1.50/$7.50 afterward. The model is available through Google AI Studio, Google Antigravity, Android Studio, Stitch, and Gemini Enterprise, as well as to Google AI Pro and Ultra subscribers in the Gemini app, AI Mode in Search, and Gemini in Google Sheets.
Google has not released a frontier-level Gemini Pro model since early 2026. ANALYSIS Three Flash releases in six weeks, with no new Pro model in that span, means Flash is now carrying Google's frontier workload in practice.
Separately, sources reported that Gemini 4 has performed well on pre-training evaluations but still needs to complete post-training1.