Google today announced Nano Banana 2 (NB2) Lite, an optimized image-generation model marketed on its API as Gemini 3.1 Flash-Lite Image. Designed to prioritize execution speed and low infrastructure expense, NB2 Lite is targeted at enterprise developers and is available immediately via Google AI Studio, the Gemini API and the Gemini Enterprise Agent Platform (GEAP).
Technical profile and performance
NB2 Lite is built on the Gemini 3.1 Flash Lite architecture to address latency and operational cost trade-offs. Google states the model can generate a standard 1k resolution image in under four seconds. The published price is a flat 0.034 USD per 1,000 images.
According to Google’s internal documentation, the model includes upgraded world knowledge to draft rough data visualizations and contextual layouts, improved character consistency to maintain identity across continuous image streams, and localized typographic rendering capabilities.
The "Lite" label imposes explicit limits: NB2 Lite supports only a 1k canvas, unlike the broader Nano Banana 2 (NB2) and Nano Banana Pro (NB Pro) lines that scale across 1k, 2k and 4k outputs. Google’s technical datasheets outline these trade-offs.
In internal benchmarks, Nano Banana 2 Lite achieved a Text to Image Elo score of 1251, compared with NB1’s 1151 and NB Pro’s 1245 on the same track. For editing tasks, the model scored 1308 for single-image editing and 1294 for multiple-image editing, figures Google positions as an optimized balance for real-time applications.
Use cases: prototyping, programmatic ads, commerce
Google frames NB2 Lite not as an artistic engine but as an invisible, high-throughput utility layer for automated workflows, rapid prototyping and programmatic applications. The target users include software engineers, programmatic advertising platforms and digital commerce applications where rapid iteration is important.
Google highlights three production environments where NB2 Lite is especially useful:
- world knowledge for quickly drafting contextual scenes or location-specific mockups;
- character consistency for storyboarding tools and digital try-ons where object fidelity must remain stable across sequential generations;
- improved text rendering that permits legible copy to be embedded directly into fast ad generation flows and checked across languages on the fly.
Google notes that while native image generation runs in the lowest-latency profiles, conditional image editing can have slightly higher response times because of secondary processing layers required to rewrite existing pixels.
Licensing and access
NB2 Lite is distributed through proprietary APIs, reflecting an enterprise-first commercial licensing approach. Unlike open-weights models that can be downloaded and run locally under open-source licenses, Google’s new models are tightly integrated into its managed cloud stack. This reduces the operational complexity of hosting hardware but ties usage to Google’s metered pricing.
Financially, NB2 Lite’s price of 0.034 USD per 1,000 images undercuts the older NB1 (0.039 USD) and is significantly cheaper than standard NB2 (0.067 USD) and NB Pro (0.134 USD). Internal notes cited by Google indicate NB2 Lite delivers approximately 60–70% of the general capability of NB2 and NB Pro while running much faster and at a fraction of the cost.
By lowering the cost of high-frequency image generation, Google is positioning NB2 Lite as a means to attract enterprise developers into its commercial platform ecosystem.
Related offerings
The NB2 Lite release coincides with the public preview of Gemini Omni Flash, a multimodal conversational video generation and editing model. While Omni Flash reflects Google’s longer-term focus on agentic video manipulation, Nano Banana 2 Lite is presented as the immediate infrastructure workhorse for high-throughput commercial applications and rapid programmatic prototyping.
