Nano Banana 2.1 vs Nano Banana Pro
Should you use Google's new Nano Banana 2.1 or Nano Banana Pro (Gemini 3 Pro Image)? In Google's model card 2.1 leads on all 10 evals, at roughly a quarter to half of Pro's per-image price. Official Elo scores, 1K/2K/4K prices and use-case picks, with live CallAI pricing.
Nano Banana 2.1
Nano Banana Pro
By Google's own model card, Nano Banana 2.1 (thinking) beats Nano Banana Pro on text-to-image, infographics, editing and character consistency, at a quarter to half the per-image price — start new projects on 2.1. If a pipeline is already tuned and happy on Pro, keep it for now and A/B a sample batch on 2.1 before switching. These are Google's own numbers with no independent replication yet.
Google model card evals and list price per image
| Benchmark | Nano Banana 2.1 | Nano Banana Pro | Winner | Notes |
|---|---|---|---|---|
| Text-to-image · Overall preference Elo | 1050 | 935 | Nano Banana 2.1 | Side-by-side human ratings |
| Infographic design Elo | 1048 | 912 | Nano Banana 2.1 | Posters and chart-style layouts |
| Infographic factuality | 0.521 | 0.265 | Nano Banana 2.1 | AutoRater: are the facts in the image right |
| General editing Elo | 1026 | 939 | Nano Banana 2.1 | Edit an uploaded photo from a prompt |
| Single-character consistency Elo | 1028 | 991 | Nano Banana 2.1 | Same person after the edit |
| Multi-character consistency Elo | 1106 | 1011 | Nano Banana 2.1 | The widest gap in the card |
| Product consistency Elo | 1024 | 965 | Nano Banana 2.1 | Product shots in new scenes |
| Multi-reference editing Elo | 1066 | 989 | Nano Banana 2.1 | Combine several reference images |
| List price per 1K / 2K image | $0.034 / $0.050 | $0.134 / $0.134 | Nano Banana 2.1 | Gemini API standard tier, as of 2026-10-23 |
| List price per 4K image | $0.113 | $0.24 | Nano Banana 2.1 | Batch API is half price |
Nano Banana 2.1 strengths & weaknesses
- Ahead of Pro on every eval in Google's model card: overall preference Elo 1050 vs 935
- Keeps people consistent across edits: multi-character consistency Elo 1106 vs 1011
- Cheaper per image: about $0.034 at 1K and $0.113 at 4K, vs $0.134 and $0.24 for Pro
- Up to 14 reference images, Google Search grounding, minimal / medium / high thinking levels
- The model card notes small text, long paragraphs and full-page text can blur at 1K
- Still slips on spatial relations (left/right) and 3D reasoning
- No caching or function calling
Nano Banana Pro strengths & weaknesses
- Built on Gemini 3 Pro with thinking support, positioned for complex, demanding images
- Out longer, so existing prompts and pipelines carry over
- Batch API at $0.067 per 1K/2K image keeps bulk jobs predictable
- Scores below Nano Banana 2.1 on every eval in Google's own card
- About 2.7× (2K) to 4× (1K) the per-image price of 2.1
Scenario recommendations
| Scenario | Pick | Why |
|---|---|---|
| E-commerce hero shots, product background swaps | Nano Banana 2.1 | Product consistency Elo 1024 vs 965, at about a quarter of the per-image price. |
| One character across many images (comics, picture books, brand mascots) | Nano Banana 2.1 | Leads on both single- and multi-character consistency, by nearly 100 Elo for multiple characters. |
| Data infographics and text-heavy posters | Nano Banana 2.1 (at 2K or above) | Infographic factuality 0.521 vs 0.265; the card warns small text blurs at 1K, so render at 2K or 4K. |
| Production pipelines already stable on Pro | Keep Nano Banana Pro for now | The same key calls both — compare a sample before switching. |
Python dual-model gateway sample
OpenAI-compatibleOne Base URL and API key — switch models with the model parameter only:
import os
from openai import OpenAI
# One CallAI key — route freely between Nano Banana 2.1 and Nano Banana Pro
client = OpenAI(
api_key="sk-live-your-callai-key",
base_url="https://api.callaiapi.com/v1"
)
def query_model(model_id: str, prompt: str):
response = client.chat.completions.create(
model=model_id,
messages=[{"role": "user", "content": prompt}],
stream=True
)
for chunk in response:
print(chunk.choices[0].delta.content or "", end="", flush=True)
print("\n" + "="*40 + "\n")
# Model 1: Nano Banana 2.1
print(">>> Output from: Nano Banana 2.1")
query_model("gemini-nano-banana-2.1", "Briefly analyze the trade-offs of this architecture")
# Model 2: Nano Banana Pro
print(">>> Output from: Nano Banana Pro")
query_model("gemini-3-pro-image", "Briefly analyze the trade-offs of this architecture")Sources
FAQ
How are Nano Banana 2.1 and Nano Banana Pro related?
Nano Banana Pro is Gemini 3 Pro Image. Nano Banana 2.1, released 2026-10-06, is a Flash-tier image model built on Gemini 3.6 Flash that Google positions as the efficient counterpart to Pro; it replaces the deprecated Nano Banana 2 (gemini-3.1-flash-image).
How much does each image cost?
Gemini API standard tier as of 2026-10-23: Nano Banana 2.1 is about $0.0336 per 1K image, $0.0504 per 2K and $0.113 per 4K; Nano Banana Pro is $0.134 per 1K/2K image and $0.24 per 4K. Batch is half price for both. The cards at the top show CallAI’s current per-token price.
Who ran these evals?
Google DeepMind, in the Nano Banana 2.1 model card: Elo from side-by-side human ratings, infographic factuality from an AutoRater. They are vendor-reported and not independently replicated.
How do I try it or call it from the API?
The CallAI Nano Banana try page gives a few free images a day. For the API, create a key, set base_url to https://api.callaiapi.com/v1, call chat/completions with model gemini-nano-banana-2.1 or gemini-3-pro-image and include image in modalities.