Google released Gemini 3.8 Flash on September 2, 2026, along with a security-focused sibling called Gemini 3.8 Flash Cyber. The new Flash model is positioned as Google’s strongest reasoning and coding model at the same speed and price as its predecessor, Gemini 3.7 Flash, and it’s already rolling out across the Gemini app, Google AI Studio, Android Studio, and the Gemini API. The Cyber variant is not publicly available: it’s gated behind a new program called Fairwind, aimed at vulnerability research and patching rather than general use.
What’s Actually New in Gemini 3.8 Flash
Google says the model works harder on complex requests by taking extra reasoning steps and calling tools iteratively before answering, rather than just responding faster. On DeepSWE v1.1, a long-horizon software engineering benchmark, Google reports 3.8 Flash outperforming most larger frontier models at solving complete engineering tasks end to end. It also posts a 54.9% score on HLE-Verified and outperforms 3.7 Flash and other frontier models on Vals’ Finance Agent benchmark and Harvey’s Legal Agent benchmark.
Pricing stays at the same introductory rate as 3.7 Flash: $0.75 per million input tokens and $3.75 per million output tokens, through December 31, 2026, after which it rises to $1.50 and $7.50. For teams already building on the Gemini API, that means a model upgrade without a pricing decision attached, at least for the next few months.
Gemini 3.8 Flash Cyber and the Fairwind Program
The Cyber variant is tuned specifically for autonomous vulnerability discovery and patching, with fewer of the guardrails that apply to the general-purpose model. Google reports it beating 3.5 Flash Cyber and larger frontier models on the CyberGym benchmark, and says Chrome’s security team measured it producing 2.6 times more correct vulnerability patches than the best commercial models tested. Wiz reported 7.5 to 9.7 percentage points higher recall at 2.3 to 5.2 times lower cost using the model.
Access runs through the Fairwind Program, which Google limits to trusted defenders: government cybersecurity authorities, healthcare and telecom operators, critical infrastructure teams, and software maintainers. Palo Alto Networks, Snowflake, and Wiz are named launch partners. There’s no self-serve signup path for the Cyber variant, so most teams will only ever touch the standard 3.8 Flash release.
Who Should Actually Care
If you’re already running agents or coding assistants on Gemini 3.7 Flash, 3.8 Flash is a drop-in upgrade worth testing before the pricing changes at year end, particularly for anything doing multi-step tool calling or longer engineering tasks. The Android Studio integration also makes this relevant for mobile teams evaluating Gemini for in-app AI features rather than just internal tooling. Deciding whether a Flash-tier model is actually the right fit, versus a heavier model or a narrower fine-tuned setup, is exactly the kind of call our AI development work involves before any code gets written, and it comes up just as often on mobile projects adding AI features as it does on backend agent work.
