Overview

• The model is live in the Gemini app for consumers and available to developers via the Gemini API, Google AI Studio, and Vertex AI.
• Key capabilities include maintaining character likeness across edits, multi-image fusion, targeted local changes from natural-language prompts, and use of Gemini’s world knowledge.
• Google lists pricing at $30 per 1 million output tokens, translating to about $0.039 per 1024×1024 image.
• All outputs carry Google’s SynthID invisible watermark, with the Gemini app also adding a visible “ai” mark, and the company points to policies that prohibit non‑consensual intimate imagery.
• The system surfaced on LMArena as “nano-banana,” where it ranked highly, and Google is expanding access through partners including OpenRouter and fal.ai while noting the preview will stabilize in the coming weeks.