Executive Summary
Google has launched the Gemini 3.1 Flash Image Preview (internally named Nano Banana 2), a new multimodal model available on its AI Gateway platform. This release focuses on enhancing image generation quality without sacrificing the speed and cost-efficiency of flash-tier models. The model introduces the ability to ground outputs using Google Image Search for greater real-world accuracy, offers configurable reasoning levels for complex prompts, and supports new image aspect ratios.
Key Takeaways
* Product Name: Gemini 3.1 Flash Image Preview (Nano Banana 2).
* Real-World Grounding: The model can use Google Image Search to retrieve live visual data, improving its ability to render lesser-known landmarks and objects accurately.
* Configurable Thinking Levels: Users can set the model's reasoning level to 'Minimal' or 'High', allowing it to perform more in-depth processing for complex prompts before generating an image.
* New Formats: Support has been added for new resolutions (512p) and aspect ratios (1:4, 1:8), expanding creative options for developers.
* Performance: The primary goal is to provide higher visual quality while maintaining the high speed and low cost characteristic of flash-tier models.
* Availability: The model is available now in preview for developers using the AI Gateway and AI SDK.
Strategic Importance
This update enhances Google's competitive offering in the AI model space by adding sophisticated, real-time grounding and reasoning capabilities to its cost-effective, high-speed model tier, appealing to developers who require both quality and efficiency.