Google launches Nano Banana 2 Lite for $0.034 per 1k images in 4

Google’s AI studio today added Nano Banana 2 Lite (NB2 Lite), a 4‑second, 1k‑pixel image generator priced at $0.034 per 1,000 images and offered through the Gemini API and Enterprise Agent Platform [2]. The low‑cost, high‑speed model is positioned as the workhorse for commercial image‑generation pipelines, while the broader Nano Banana 2 family remains the default in Google’s Gemini chatbot [1].

Speed and cost versus the legacy line

NB2 Lite builds on the Gemini 3.1 Flash Lite architecture, trimming the resolution to a fixed 1k canvas but cutting generation time to under four seconds [2]. In internal benchmarks the model posted a Text‑to‑Image Elo score of 1251, beating the original Nano Banana (1151) and edging out the larger Nano Banana Pro (1245) [2]. By contrast, the older NB1 model cost $0.039 per 1,000 images, so NB2 Lite reduces per‑image spend by roughly 14 % [2].

Enterprise use cases and competitive context

Google markets NB2 Lite as a “high‑throughput utility layer” for programmatic ad creation, rapid A/B testing, and automated asset generation, where speed and cost matter more than multi‑resolution flexibility [2]. The model’s upgraded world knowledge and character‑consistency features aim to keep identity stable across sequential edits—a known pain point for marketers and e‑commerce platforms [2]. Competitors such as Krea’s “2 Turbo” model claim two‑second generation times and open‑weight licensing, but Google’s offering benefits from tight integration with its broader Workplace and AI ecosystem [2].

Access and early impressions

Developers can call NB2 Lite through the Gemini API, AI Studio, or the Gemini Enterprise Agent Platform, all of which bundle the model with Google’s managed cloud services [2]. Hands‑on testing via the Gemini chatbot shows the model can pull real‑time web data for infographic creation, though accuracy can lag (e.g., outdated weather data) [1]. Google also watermarks outputs to flag AI‑generated images, though the watermarks can be subtle enough to miss casual viewers [1].

What to watch

• Rollout milestones – Google’s upcoming public preview of Gemini Omni Flash, a multimodal video generation model, may extend the same speed‑first philosophy to moving images.
• Pricing adjustments – Any shift in the $0.034 per 1,000‑image rate could affect enterprise adoption, especially if competitors tighten their cost structures.
• Competitive responses – Krea’s open‑license Turbo model and other emerging low‑latency generators could pressure Google to further lower latency or open model weights.

Google’s NB2 Lite demonstrates that enterprise‑grade image generation can now be both cheap and fast, narrowing the gap between large‑scale AI labs and commercial developers while raising the bar for how quickly visual assets can be produced at scale.