Google quietly shipped two new Gemini models today (Nano Banana 2 Lite + Omni Flash). Excited??

Google recently announced and pushed two new models into the Gemini API. For informational purposes, I am breaking down what's confirmed vs. what's just early testing chatter so far.

Nano Banana 2 Lite (gemini-3.1-flash-lite-image):

• Text-to-image in around 4 seconds.

• $0.034 per 1K-resolution image.

• Positioned as the direct replacement for the original Nano Banana. (gemini-2.5-flash-image)

• Rolling out across AI Studio, the Gemini API, Gemini Enterprise, and consumer surfaces. (Search, Gemini app, Photos, Flow)

Gemini Omni Flash (gemini-omni-flash-preview):

• Public preview, AI Studio + Gemini API.

• Conversational editing: now you can refine a generated video with plain language instead of re-prompting from scratch.

• Multimodal input (text/image/video combined) to keep a scene consistent.

• $0.10/sec of video output, same rate as Veo 3.1 Fast.

• Confirmed limitations: Capped at 10 sec of generations for now, no audio reference uploads or scene extension yet in the API, character consistency across scene changes still has rough edges.

What's actually more interesting than the specs is that these two are clearly meant to be chained. Generate the image fast and cheaply with a lighter version, then feed it into Omni Flash to animate it. Google's demo apps (Anywhere, Space Lift, Omni Product Studio) are all built around that exact pipeline.

Early hands-on reports are mixed. A few people testing the image model are seeing real speed gains, roughly half the generation time on basic tasks, but anything detail-heavy (blurry text, fine classification) still needs the full model. Omni Flash reactions lean more toward a "lateral move" than an upgrade; some side-by-sides found output quality close to the previous flash model, with the latency gain landing closer to ~200ms than a real leap.