Google DeepMind ships Nano Banana 2 Lite, fastest image model at $0.034 per 1K images

Google DeepMind released Nano Banana 2 Lite (gemini-3.1-flash-lite-image), its fastest and most cost-efficient image generation model, designed for high-throughput, speed-constrained developer workflows. The model delivers text-to-image outputs in 4 seconds at $0.034 per 1K-resolution image, positioning it as a direct replacement for the original Nano Banana (gemini-2.5-flash-image). It is available immediately in Google AI Studio, Gemini API, and Gemini Enterprise Agent Platform, with rollout to Google consumer surfaces including AI Mode in Search, Gemini app, NotebookLM, and Google Photos.

Despite prioritizing ultra-low latency, Nano Banana 2 Lite retains reliable prompt adherence, strong character consistency, and legible in-image text rendering. The model is part of a tiered Nano Banana family: Nano Banana 2 Lite for speed and cost, Nano Banana 2 as a generalist workhorse balancing performance and cost, and Nano Banana Pro for complex professional use cases. Google also announced Gemini Omni Flash for video generation and editing, priced at $0.10 per second of output, supporting conversational editing, multimodal referencing, and real-world knowledge grounding.

For developers and product teams, Nano Banana 2 Lite enables rapid visual iteration at 4-second latency—critical for interactive prototyping, real-time drafting, and cost-constrained production pipelines. The aggressive pricing and speed positioning it directly against open-source and competing commercial models, making it a strong play for high-volume image-generation use cases where developer velocity and per-request cost matter more than peak quality.