Google launches Nano Banana 2 Lite for lightning-fast, budget-friendly image generation

Google DeepMind has introduced Nano Banana 2 Lite, a stripped-down image generation model designed to deliver rapid outputs at minimal cost. Part of the Gemini 3.1 family (formally Gemini 3.1 Flash Lite Image), the new model addresses a long-standing tension in AI imagery: choosing between speed, quality, and expense.

Speed and cost advantages#

The standout feature is velocity. Nano Banana 2 Lite can transform text prompts into complete images in roughly 4 seconds using its default low-thinking mode—roughly five times faster than the standard Nano Banana, which requires about 20 seconds. This efficiency translates directly to developer wallets: Google prices the model at an average of $0.034 per 1,000 images via its API. By contrast, the original Nano Banana costs twice as much, and the more capable Nano Banana Pro reaches $12 per million output tokens.

Trade-offs in quality#

Speed comes with compromises. While Arena.ai benchmarks show user ratings for Nano Banana 2 Lite nearly matching its non-Lite variants, the model struggles with finesse in specific areas. Text rendering remains unreliable, especially at small sizes, and infographics frequently contain factual errors. Character and human depictions may lack consistency between successive generations—a lingering issue in economical image models.

The use case#

Google positions Nano Banana 2 Lite as a tool for exploratory work and rapid prototyping, where quick iteration matters more than flawless polish. Developers can access the model across the Google ecosystem, including via API and Gemini applications.