Gemini 3.1 Flash Image (Nano Banana 2)
Key facts · verified Oct 13, 2026
Gemini 3.1 Flash Image (Nano Banana 2) is a Google image generation model released on 28 May 2026. Status: Deprecated (no shutdown date announced); replacement: gemini-nano-banana-2.1. Pricing: $0.045 (0.5K) / $0.067 (1K) / $0.101 (2K) / $0.151 (4K) per image. Source: ai.google.dev.
Nano Banana 2 is Google's name for Gemini 3.1 Flash Image (gemini-3.1-flash-image), an image-generation model available via the Gemini API. A preview version was released in February 2026 and the stable model on May 28, 2026. Google prices it per output image, with the rate rising with resolution from 0.5K to 4K. Google's image-generation docs recommend Nano Banana 2.1 for all new projects.
$0.045 (0.5K) / $0.067 (1K) / $0.101 (2K) / $0.151 (4K) per image
Not published
Not rated
Not ranked
Key Features
Full Specifications
$0.045 (0.5K) / $0.067 (1K) / $0.101 (2K) / $0.151 (4K) per image | |
|---|---|
| Identity | |
| Developer | |
| Released | May 2026 |
| Status | Deprecated (no shutdown date announced); replacement: gemini-nano-banana-2.1 |
| Licence | Proprietary |
| Self-hostable | No |
| Cost | |
| Price per image | Standard: $0.045 (0.5K/512px), $0.067 (1K), $0.101 (2K), $0.151 (4K); Batch API (50% off): $0.022/$0.034/$0.050/$0.076 |
| Cheapest plan | No subscription plan - pay-as-you-go via the Gemini API only, billed per output image (see Price per image) |
| Free tier | Not offered via the Gemini API (free tier explicitly excludes image input/output pricing rows); available at no cost, subject to shared compute-based usage limits, in the consumer Gemini app on every tier including "Without a Google AI plan" |
| Capacity | |
| Max resolution | Up to 4096x4096px (4K, 1:1 ratio); other ratios scale within the same per-tier token budget, e.g. 12288x1536 at 8:1. Also 512x512 (0.5K/512px), 1024x1024 (1K, default), 2048x2048 (2K) |
| Aspect ratios | 14 ratios: 1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9, 21:9 |
| Images per request | 1 per interaction (single output_image; no batch/count parameter - generate more via multi-turn conversation or separate calls) |
| Capability | |
| Inpainting | Yes - conversational/semantic masking ("define a mask to edit a specific part of an image while leaving the rest untouched") |
| Image-to-image | Yes (text-and-image-to-image editing, up to 14 total reference images: up to 10 high-fidelity objects + up to 4 character-consistency images) |
| Text rendering | Yes - advanced text rendering, improved i18n/localized text in the 3.1 update |
| Style controls | Not offered as a dedicated feature (no style-reference-image parameter - exclusive to Gemini 3 Pro Image, up to 3 images); restyling only via freeform text prompting |
| Character consistency | Yes - up to 4 reference images of characters to maintain consistency (vs. up to 5 for Gemini 3 Pro Image) |
| Rights | |
| Commercial rights | Yes - Google claims no ownership over generated content; no explicit prohibition on commercial use, subject to the Prohibited Use Policy and applicable law |
| Training-data opt-out | Yes on paid tier - "Google doesn't use your prompts...or responses to improve our products" on Paid Services; no opt-out on free/unpaid tier or AI Studio, where content "improve[s]...Google products and services" by default (human reviewers may read/annotate) |
| Access | |
| API | Yes (Gemini API) |
| Chat app | Yes - default image-generation engine across the Gemini app's Fast/Thinking/Pro modes, on every subscription tier including free |
| Cloud marketplaces | Not offered |
| Fine-tuning | Not offered (no Gemini API model currently supports fine-tuning) |