Google Begins Internal Testing of Gemini 3.8 Flash Model
Google has reportedly begun internal testing of its Gemini 3.8 Flash Preview model on its internal programming platform, Jetski. This rapid update comes shortly after the release of Gemini 3.7 Flash, reflecting Google's push for a near-monthly model release cycle. As Google struggles to match competitors' top-tier frontier models, it is prioritizing the faster and cheaper Flash series to power AI agents like Gemini Spark. These agents consume tokens rapidly, making cost-efficiency and speed crucial for enterprise adoption. Early feedback from testers indicates Gemini 3.8 Flash performs noticeably better than version 3.7 Flash, though testing is still in its early stages. The focus on Flash models aligns with the industry-wide race to build autonomous consumer agents, such as Meta's upcoming Hatch agent.
## BACKGROUND
Google's Gemini model family includes different tiers, with Pro designed for complex tasks and Flash optimized for speed and cost-efficiency. AI agents, such as Google's Gemini Spark, are autonomous assistants designed to perform multi-step tasks across various applications, requiring high-frequency model calls.