Nano Banana 2.1: release, price, vs Nano Banana 2
Google released the Nano Banana 2.1 image model on 6 October 2026: 4K, 14 references, $0.0336 per 1K image on the API, half of Nano Banana 2's price.
Google released Nano Banana 2.1 on 6 October 2026, generally available as gemini-nano-banana-2.1 in the Gemini API. It updates Nano Banana 2 with better quality, text rendering and character consistency, and outputs up to 4K. A 1K image lists at $0.0336, half Nano Banana 2's price. Versely lists it from 4 credits (16 October 2026).
Key facts
| Item | Detail |
|---|---|
| Released | 6 October 2026, generally available |
| Model ID | gemini-nano-banana-2.1, based on Gemini 3.6 Flash |
| Output | 1K (default), 2K, 4K; 14 aspect ratios, 1:8 to 8:1 |
| References | Up to 14: up to 10 objects and 4 characters |
| API price | $0.0336 (1K), $0.0504 (2K), $0.113 (4K) per image |
| Where | Gemini app, AI Studio, Gemini API, AI Mode in Search, Google Ads, Flow, Stitch |
| Nano Banana 2 | Deprecated; no shutdown date announced |
| On Versely | 4 credits at SD/HD, 7 at 4K (16 October 2026) |
Sources, all read on 16 October 2026: Google's changelog, deprecations page, model page, image guide and pricing page; DeepMind's launch page and model card; Artificial Analysis' text-to-image and editing boards. The model card lists channels, not dates: the Gemini app release notes had no 2.1 entry on 16 October, so check your own model picker.
What is new versus Nano Banana 2
Google calls 2.1 "an update to Nano Banana 2" with "Flash-level speed and cost efficiency". Side by side:
| Item | Nano Banana 2 | Nano Banana 2.1 |
|---|---|---|
| Sizes | 0.5K, 1K, 2K, 4K | 1K, 2K, 4K |
| Thinking levels (default) | minimal, high (minimal) | minimal, medium, high (medium) |
| Text-to-image Elo | 1139 | 1172 |
| Editing Elo | 1124 | 1152 |
Elo is Artificial Analysis' blind-vote score, read 16 October 2026. It ranks 2.1 fourth for text to image (range 3 to 6, 6,003 votes) and third for editing (range 3 to 5, 6,684 votes). GPT Image 2.5 Sunburst and Flare lead both boards.
Quality and text. Google lists better visual quality and realism at all three sizes, prompt adherence, multi-turn character consistency, and "enhanced text rendering and infographic layout accuracy". In the model card's human tests, with thinking on, 2.1 scores 1050 for overall preference against 990 for Nano Banana 2, and 1106 against 978 for multi-character consistency. With thinking off, overall preference drops to 1015. Those are Google's numbers, not Versely's tests.
Wide frames. The 1:4, 4:1, 1:8 and 8:1 ratios are not new: Google's Nano Banana 2 page already lists them. What changes is that Google "fixed tiling artifacts" on them at 2K and 4K. At 4K an 8:1 frame is 12288 × 1536 pixels.
Mask-based editing. DeepMind lists it as a headline gain. Its example is a text prompt ("mask the object inside the sketch and place in a completely different environment"), and the API guide documents masking only as "semantic masking", defined "conversationally" in the prompt. The model card scores "Mask/Ink-Based Editing" at 1049 against 965 for Nano Banana 2, and lists "ink persistence in masked/doodle based editing" as a known limitation.
Unchanged. Per Google's guide, reference limits, search grounding and video input match Nano Banana 2.
Subject counts differ. The API docs say up to 4 characters and 10 objects; DeepMind says "up to five characters" and "up to fourteen objects". Plan around the API numbers.
API price against Nano Banana 2
| Per image | Nano Banana 2 | Nano Banana 2.1 |
|---|---|---|
| 1K | $0.067 | $0.0336 |
| 2K | $0.101 | $0.0504 |
| 4K | $0.151 | $0.113 |
| Input, per million tokens | $0.50 | $1.50 |
Google's pricing page, updated 15 October 2026, bills 2.1's image output at $30 per million tokens against $60, which halves a 1K or 2K image. At 4K Google counts 3,780 tokens against 2,520, so the cut is about 25%. Input costs three times as much, thinking text bills at $7.50 per million on top, the Batch API halves every figure, and there is no free API tier.
One mismatch: the image guide's size table shows 2,520 tokens at 4K, which would be $0.0756, and AI Weekly (7 October 2026) quotes that. The pricing page and Origami (7 October) say $0.113. Read a test call's usage metadata before you budget 4K at volume.
Is Nano Banana 2 being shut down?
Google has announced no shutdown date, as of 16 October 2026. Its changelog entry of 6 October said gemini-3.1-flash-image "is deprecated and will be shut down on October 29, 2026", as an Internet Archive copy from 16:21 UTC that day shows. A copy from 12:10 UTC on 7 October and the live changelog both read "deprecated (no shutdown date announced)", and the deprecations page names gemini-nano-banana-2.1 as the replacement. The 29 October date in early coverage matches the removed wording.
Google Cloud's lifecycle page for its Gemini Enterprise Agent Platform, read the same day, gives Nano Banana 2 a retirement date of "May 28, 2027 or later" and says dates won't move earlier. Google's guide calls Nano Banana 2 "previous-generation" and recommends 2.1 for all new projects, so plan a migration for old code.
Who it is for, and who should wait
Good fit:
- Posters, thumbnails, infographics and mockups where text and layout matter. Google lists both as gains, and 2.1 ranks fourth for text to image on Artificial Analysis.
- Series that keep the same characters across frames. Multi-character consistency is the model card's widest margin, 1106 against 978.
- Nano Banana 2 users, since both rows start at 4 credits on Versely, and API teams moving off the deprecated ID once they have re-run their cost estimate.
Test first, or wait, if:
- You need more than 10 references on Versely, where Google's API takes 14, or bulk 4K there, where the Nano Banana 2 row is a flat 4 credits against 7.
- Your frames carry small text or long paragraphs. The model card lists both as weaknesses, with small text "often blurry in 1k". Try 2K.
- Numbers and labels must be right. The card says 2.1 is "still limited in advanced capabilities with world knowledge, 3D reasoning and factuality".
- You need a painted mask. The Versely row lists no mask input, and Google flags "ink persistence".
- You rely on Nano Banana Pro's listed strengths: world knowledge, localization, brand consistency. 2.1 still beats Pro in Google's preference test (1050 against 935) and on Artificial Analysis' text-to-image board (1172 against 1114), so test your own frames.
How to use it on Versely
Nano Banana 2.1 is one row on Versely, covering text to image, image to image and editing; there is no separate Edit row. As listed on 16 October 2026 it costs 4 credits at SD/HD, currently marked down 50% from 7, and 7 credits at 4K, down from 13, so a job runs 4 to 7 credits by output resolution, shown before you generate. The discount can change.
The row offers 1K, 2K and 4K output, 14 aspect ratios from 8:1 to 1:8 plus auto, up to 10 reference images and an optional starting image. It lists no mask input, search grounding or thinking-level setting. The GPT Image 2.5 edit rows, such as GPT Image 2.5 Flare Edit, list an optional mask.
- Open the text to image tool and choose Nano Banana 2.1 from the model list.
- Prompt the subject, the shot, the light and the job of the frame. The Nano Banana 2 prompting guide covers the technique, written for the earlier model.
- Choose ratio and quality, and try 2K or 4K for text-heavy frames.
- To edit, attach the image and describe one change. Google's template: "Using the provided image, change only the [specific element] to [new element/description]. Keep everything else in the image exactly the same, preserving the original style, lighting, and composition."
Nano Banana 2 stays at a flat 4 credits whatever the resolution, with 14 references against 10 here.
FAQ
When was Nano Banana 2.1 released?
Google released Nano Banana 2.1 on 6 October 2026, per its Gemini API changelog. DeepMind published the model card the same day.
How much does the Nano Banana 2.1 API cost?
Google lists $0.0336 per 1K image, $0.0504 per 2K and $0.113 per 4K on the standard Gemini API, plus $1.50 per million input tokens, as read on 16 October 2026.
Is Nano Banana 2 being shut down?
Google has announced no shutdown date for Nano Banana 2. Its changelog named 29 October 2026 on 6 October, then changed to "no shutdown date announced", which it still says on 16 October 2026.
Is Nano Banana 2.1 better than Nano Banana 2?
On the evidence so far, yes. Artificial Analysis' votes put it ahead, 1172 against 1139 for text to image and 1152 against 1124 for editing, and Google's own tests agree. The API costs half as much at 1K and 2K, but three times as much per input token.
Can I use Nano Banana 2.1 on Versely?
Yes. As listed on 16 October 2026, it costs 4 credits at SD/HD and 7 at 4K, with up to 10 reference images.