Gemini 3.8 Flash holds token prices as Cyber access stays limited

Google has launched Gemini 3.8 Flash at the same introductory API price it used for Gemini 3.7 Flash, while keeping its cyber-focused version behind a separate trusted-access program.

The confirmed price point is $0.75 per million input tokens and $3.75 per million output tokens until December 31, 2026. Google says the standard 3.8 Flash model improves on 3.7 Flash in software engineering, agentic tasks and multi-step reasoning. The practical caveat is that stronger reasoning can still raise the total bill when an application asks the model to take more steps or use more tokens.

Same unit price, different workload risk

Google frames 3.8 Flash as a workhorse model for long-horizon coding and autonomous agents. It points to gains on software-engineering and professional-domain benchmarks, but those are Google-run or Google-cited claims. The firmer product change for developers is simpler: the newer Flash model is available through Google AI Studio, Android Studio, Gemini Enterprise, the Gemini app, AI Mode in Search and Gemini in Google Sheets.

For teams already testing 3.7 Flash, the pricing detail matters because a model can become more capable without becoming easier to budget. Google says 3.8 Flash may execute extra reasoning steps and call tools iteratively on complex tasks. That makes lower effort levels and workload measurement part of the upgrade decision, not an afterthought.

Cyber is treated differently

Gemini 3.8 Flash Cyber is not being rolled out like a normal public model. Google's Fairwind Program gives prioritized access to government authorities, critical infrastructure operators, software maintainers and other trusted partners. Google says the cyber model is aimed at vulnerability discovery and patching, with safeguards and operational standards for participating organizations.

That split shows where model releases are heading. A general model can be pushed into consumer, enterprise and developer channels quickly. A stronger cyber model may need a narrower access path because the same capability that helps defenders find and fix flaws can also increase misuse risk if released broadly.

Google also says introductory Gemini 3.8 Flash pricing expires on January 1, 2027, when API rates rise to $1.50 per million input tokens and $7.50 per million output tokens. That gives developers a short window to test whether the added reasoning is worth the eventual higher unit cost.

For builders, the useful test is measurement: compare successful task completion, latency, token use and policy requirements before replacing an existing Flash deployment.