Nano Banana 2 (Gemini 3.1 Flash Image)

2026-02-26 · Publisher : Google · Domain : Artificial Intelligence

Google DeepMind launches Nano Banana 2 (Gemini 3.1 Flash Image), an image model that brings the capabilities of Nano Banana Pro to the speed of Gemini Flash, from 512 px to 4K.

What changes

  • World knowledge and web search to depict specific subjects, infographics and diagrams.
  • Rendering and translation of text within the image.
  • Subject consistency: up to five characters and 14 objects in a single workflow.
  • Better instruction following, resolutions from 512 px to 4K, several aspect ratios.
  • SynthID marking complemented by C2PA Content Credentials.

Google DeepMind launches Nano Banana 2 (Gemini 3.1 Flash Image), an image model that brings the capabilities of Nano Banana Pro to the speed of Gemini Flash, from 512 px to 4K. ### What changes - World knowledge and web search to depict specific subjects, infographics and diagrams. - Rendering and translation of text within the image. - Subject consistency: up to five characters and 14 objects in a single workflow. - Better instruction following, resolutions from 512 px to 4K, several aspect ratios. - SynthID marking complemented by C2PA Content Credentials. ### Specifications - Inputs: text and images; outputs: image (4K) and text. - Context: up to 1 million tokens of input, 64,000-token text output (model card). - Availability: Gemini app, Search (AI Mode, Lens), AI Studio and API in preview, Vertex AI, Flow, Google Ads. - API identifier: gemini-3.1-flash-image (GA version released on May 28, 2026). Knowledge cutoff: January 2025. ### Pricing | Item | Standard | Batch | |---|---|---| | Input (per 1M tokens) | USD 0.50 | USD 0.25 | | Text output (per 1M tokens) | USD 3.00 | USD 1.50 | | Image output (per 1M tokens) | USD 60.00 | USD 30.00 | Pricing recorded from the current pricing page. ### Benchmarks | Benchmark (scores published by the vendor, Elo) | Nano Banana 2 | Nano Banana Pro | GPT-Image 1.5 | |---|---|---|---| | Overall preference (GenAI-Bench) | 1073 | 1021 | 1047 | | Visual quality (GenAI-Bench) | 1129 | 1043 | 975 | | Infographics (factuality) | 1074 | 1102 | 985 | Version with thinking and search: 1079, 1140 and 1114 respectively. ### Key takeaways A fast image model that becomes the default in the Gemini app and in Flow; Nano Banana Pro remains recommended for tasks requiring the highest factual fidelity.

Links

Tags : Image, Google, Gemini, Nano Banana, Génératif