Ars Technica article
Google releases Gemini 3.8 Flash, its third Flash model in six weeks
https://arstechnica.com/ai/2026/09/google-releases-gemini-3-8-flash-its-third-flash-model-in-six-weeks/
Part of the content is translated by Gemini Flash.
Google hasn't released a Frontier-class Gemini Pro AI model since early 2026, but it continues to unveil new Flash variants. Now Google has announced its third Flash model in just six weeks, further increasing the likelihood that we won't see the initially promised Gemini 3.5 Pro. However, Google states that Gemini 3.8 Flash is its highest-performing inference and coding model, so there shouldn't be a major issue.
Gemini 3.8 Flash comes in two variants. First, the general Flash is described by Google as a 'workhorse' model suitable for everything from agent tasks to software development. Second, Gemini 3.8 Flash Cyber is tuned for vulnerability detection and mitigation while operating on the same foundation.
For developers, Google has offered the same terms as when it launched 3.7 Flash just a few weeks ago. API access to this model is available at a 'launch discount' of $0.75 per million input tokens and $3.75 per million output tokens until the end of the year. The regular price will be $1.50/$7.50, but it's highly likely that a new model will arrive long before any price changes. Considering that other AI research labs have recently lowered token prices to maintain corporate customer engagement, Google appears to view low pricing as an inevitable choice.
[Rest of the content omitted]
https://x.com/Google/status/2095193918686077027
