Google Debuts Third Flash Model in Six Weeks as Premium Gemini Releases Stall

Google is leaning heavily on rapid iterations of its lightweight Gemini Flash models, having unveiled its third such release in just six weeks. The latest addition—Gemini 3.8 Flash—aims to strengthen Google's competitive footing in reasoning and code generation, even as the company appears to have shelved the anticipated Gemini 3.5 Pro flagship.

Two Flavors of Flash#

The new release comes in two configurations. The base variant serves as a general-purpose "workhorse" designed to handle everything from AI agent work to programming tasks. A specialized sibling, Gemini 3.8 Flash Cyber, shares the same underlying architecture but has been fine-tuned specifically to detect and counter security vulnerabilities.

Pricing and Availability#

Developers can access both models through an introductory pricing window that extends through year-end: $0.75 per million input tokens and $3.75 per million output tokens. Standard pricing will rise to $1.50 and $7.50 respectively, though Google's track record suggests newer models will likely arrive before any price increases take effect. The aggressive promotional rates reflect broader industry pressure—rival labs have slashed token costs to maintain developer engagement amid growing skepticism around AI's immediate business value.

Performance Claims#

Google has published benchmark results positioning Gemini 3.8 Flash as competitive with larger, pricier models. Against its predecessor Gemini 3.7 Flash (released just weeks prior), the improvements are modest across most tasks but notably stronger in coding evaluations.