whatmodel.sh / image / Nano Banana Pro (Gemini 3 Pro Image)

Nano Banana Pro (Gemini 3 Pro Image)

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Nano Banana Pro (Gemini 3 Pro Image) is a closed-weights image model from Google, released November 2025. Licence: proprietary.

75.8±1.1#9 of 144 image modelsfrom 20 benchmark results across 16 sources

Score within the LMArena text-to-image comparison set only.

$ how it compares

The image models ranked either side of Nano Banana Pro (Gemini 3 Pro Image), out of 197 with a score, then more from Google.

#modelscorewhereRAMreleased
8Muse Image Meta76±2api $0.010/image2026-07
9Nano Banana 2 (Gemini 3.1 Flash Image) Google76±1api $0.067/image2026-02
10Nano Banana Pro (Gemini 3 Pro Image) Google76±1api $0.13/image2025-11
11Qwen-Image-3.0-Pro Alibaba75±2api $0.040/image2026-07
12Seedream 5.0 Pro ByteDance Seed75±2api $0.050/image2026-07
more from Google
5Nano Banana 2.1 Google81±2api $0.034/image2026-10
16Nano Banana 2 Lite (Gemini 3.1 Flash-Lite Image) Google73±2api $0.034/image2026-06
39Nano Banana (Gemini 2.5 Flash Image) Google62±1api $0.039/image2025-08

see all image models · find the ones that fit your machine

$ what it's good at

text to image▂▅▅▃▄▄█▃▃▂▁▁ #12 of 126 · 42% of its score · 8 tests

Trails Reve 2.0, Seedream 5.0 Pro, Qwen-Image-3.0-Pro here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
LMArena text-to-image1234.1 ✎▂▂▃▄▇█▄▃▄▂·▂ beats 60/75avg 1027▇▇▇▇▇▇▇▇ 19%typical
GenExam (relaxed overall, GPT-5 judge)93.0% ✎▄█▂▂▂▂▆▄·▂·▃ beats 16/16hard · avg 26%▇▇▇▇▇▇ 15%sharp
TIIF-Bench v3 (short prompts, TIIF-Evaluator)93.1% ✎▂█····▄▄▄··▃ beats 11/11easy · avg 79%▇ 2.4%maxed out
FEPBench instruction faithfulness62.9% ✎▄·█···▄▄▄▄▄▄ beats 7/8medium · avg 36%▇ 2.4%blunt · at its level
Qwen-Image-Bench (Q-Judger, v1)59.5% ✎█▅▅·▅▅█·█··▃ beats 14/17medium · avg 46%▇ 2.0%blunt · at its level
HPSv3 score of Cosmos 3 report text-to-image evaluation images11.7800 ✎▄···█··▄·▄·▄ beats 4/5avg 11.076▇ 1.2%blunt
CVTG-102ch PNED (Chinese character rendering, Cosmos 3 report)76.4% ✎▄·······▄█▄▄ beats 5/5medium · avg 60%▇ 0.3%blunt · noisy
CVTG-500L PNED (English long-prompt text rendering, Cosmos 3 report)71.8% ✎█······█████ beats 0/5easy · avg 85%▇ 0.0%blunt
image to image / editing▁▁·▃▂▃█▃▃▄▁▁ #7 of 62 · 26% of its score · 5 tests

Beats Reve 2.1, Nano Banana 2 (Gemini 3.1 Flash Image) here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
Edit-Compass (EN, overall)79.8% ✎▃▃▂·▃▆▂█·▃·▃ beats 20/20hard · avg 45%▇▇▇▇▇ 12%at its level
Arena image edit1386.0 ✎▂▁▁▂▆▇█▄▆▅▁▂ beats 42/53avg 1153▇▇▇ 7.3%typical
GEBench GUI state generation (English GE-Score)61.2% ✎▄▄····▄▄█▄·█ beats 7/8hard · avg 34%▇ 3.3%at its level
GEditBench v2 (overall Elo)1096.0 ✎▃▂····▂█▂▃▂▃ beats 14/14avg 956▇ 2.9%blunt
TECCI human overall success rate (IRCS+GGIS) lab22.3% ✎▄········▄█▄ beats 4/4very hard · avg 7.1%▇ 1.1%typical
other tests7 not in a task · 31% of its score
testresultvs the others in its rundifficultycountswhy
InfoEdit infographic editing (average success rate)41.6% ✎▄··▄█····▄▄▄ beats 5/6very hard · avg 1.4%▇▇▇▇▇▇ 16%sharp · at its level
VVRBench-Challenge accuracy (720 harder verifiable tasks)4.6% ✎▄·▄···█·▄▄·▄ beats 3/6very hard · avg 0.6%▇▇ 5.1%sharp · too hard for it
ChartSync chart editing (Overall)87.8% ✎▂··▂█▂▄▂▂·▂▃ beats 12/12medium · avg 41%▇▇ 4.0%typical
Rapidata Benchmark.ai image benchmark (overall Elo)1071.6 ✎▁▁▁▄▂█▂▃▁▂▁▁ beats 26/34avg 875▇ 3.0%typical
UniGenBench + Physical-AI prompts, All 1,170 (Cosmos 3 report)90.7% ✎▄···█··▄···█ beats 4/5easy · avg 78%▇ 1.3%maxed out
ViGoR-Bench zero-shot visual reasoning, image models (Result Avg)68.4% ✎▂···▂·▂█▆▅·▃ beats 15/15hard · avg 31%▇ 1.2%at its level
Knowledge Visualization T2I (Brief Caption, English overall)34.4% ✎▅·▃·▃··▃·█▃▃ beats 6/9hard · avg 28%▇ 0.7%blunt · at its level

$ where to run it

Through an API

api · $0.13/image · ai.google.dev ↗ · price sourceGoogle Gemini Developer API Standard paid tier; 1K generated image output only. Input/text/thinking/tool charges additional. Product: Gemini Developer API. API model: gemini-3-pro-image. Settings: tier=Standard paid, resolution=1K, batch=no, n=1. Price basis: provider-published approximate image OUTPUT cost. Additional/excluded charges: Input tokens, text/thinking output tokens and optional tools. Currency: USD.

$ how the score was worked out · LMArena leaderboard dataset 26%, InfoEdit: Probing Global Layo… 16%, GenExam: A Multidisciplinary… 15% and 13 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
LMArena leaderboard dataset2▇▇▇▇▇▇▇▇ 26%CC BY 4.0
InfoEdit: Probing Global Layout Reasoning in Infographic Editing (arXiv:2609.33286v1)1▇▇▇▇▇ 16%CC-BY-4.0
GenExam: A Multidisciplinary Text-to-Image Exam (arXiv v5)1▇▇▇▇ 15%CC-BY-4.0
Edit-Compass & EditReward-Compass: A Unified Benchmark for Image Editing and Reward Modeling1▇▇▇▇ 12%CC-BY-4.0
Verifiable Visual Rewards Transfer from Synthetic Scenes to Natural Prompts (VVRBench), arXiv:2609.35641v11▇▇ 5%CC-BY-4.0
ChartSync: A Benchmark for Visuo-Logical Cascading Chart Editing (arXiv:2607.10301v1)1▇ 4%CC-BY-4.0
GEBench: Benchmarking Image Generation Models as GUI Environments1▇ 3%CC-BY-4.0
Rapidata/text-to-image-human-preference-6M (Benchmark.ai Image Benchmark) published dataset-card rankings1▇ 3%CC-BY-4.0
GEditBench v2: A Human-Aligned Benchmark for General Image Editing1▇ 3%CC-BY-4.0
Cosmos 3: Omnimodal World Models for Physical AI (arXiv:2606.02800v4)4▇ 3%CC-BY-4.0
TIIF-Bench: How Does Your T2I Model Follow Your Instructions? (arXiv v3)1▇ 2%CC-BY-4.0
Faithful, Enriched, and Precise: Benchmarking Natural-Science Illustration Generation by T2I models (FEPBench)1▇ 2%CC0-1.0
Qwen-Image-Bench: From Generation to Creation in Text-to-Image Evaluation1▇ 2%CC-BY-4.0
ViGoR-Bench: How Far Are Visual Generative Models From Zero-Shot Visual Reasoners? (arXiv:2603.25823v1)1▇ 1%CC-BY-4.0
TECCI: Tricky Edits of Collected and Curated Images · lab-reported1▇ 1%CC-BY-4.0
Knowledge Visualization: A Benchmark and Method for Knowledge-Intensive Text-to-Image Generation (arXiv:2604.22302v1)1▇ 1%CC-BY-4.0
prior · pull toward the middle▇ 1%
weighted average= 75.8

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json