Image modelActive
Nano Banana 2.1
- Released
- October 6, 2026
- Data date
- October 7, 2026
- Model type
- Maximum resolution
Nano Banana 2.1 improves prompt adherence, character consistency across editing turns, and text rendering over Nano Banana 2, according to Google. A workflow can use up to 14 reference images.
Specifications and access
| Specification | Value and source |
|---|---|
| Model type | Proprietary modelSource |
| Access | Gemini API and Google AI StudioSource |
| Maximum resolution | Up to 4KSource |
| Native audio | No audio generationSource |
| Generator | gemini-nano-banana-2.1Source |
| API model ID | gemini-nano-banana-2.1Source |
| Release | Generally available since October 6, 2026Source |
| Capabilities | Image generation and conversational editing; improved prompt adherence, character consistency, and text renderingSource |
| Reference images | Up to 14 reference images in one workflowSource |
| Input modalities | Text, images, video, and PDFSource |
| Output modalities | Images and textSource |
Page 1 of 5
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
BudgetPixel Arena: failed image attempts (Show measurement, test conditions, and source)
- Source value
- 0
- Score
- 0
- Metric
- Attempts that returned no image
- Unit
- %
- Benchmark version
- BudgetPixel AI Image Arena prompt suite, default settings, 2026-10-07 model-page snapshot
- Category
- generation-reliability
- Direction
- Lower is better
- Sample
- 68
- Source type
- independent-evaluator
- Evaluator
- BudgetPixel AI Image Arena model page
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Independent evaluation. BudgetPixel classifies failed generation attempts separately from content-policy refusals. Its own default-setting suite reports 0.0% failures across 68 generation attempts and lists one refused prompt outside that failure rate.
- Context
- This is a small, platform-specific sample. A refused prompt is not counted as a failed image attempt, so the refusal is not included in the 0.0% value.
Arena Image Edit, Multi-Image: 3D Imaging & Modeling (Show measurement, test conditions, and source)
- Source value
- 1471
- Score
- 1471
- Metric
- Bradley-Terry Arena score
- Unit
- points
- Benchmark version
- Arena public leaderboard (live page)
- Category
- image-editing
- Direction
- Higher is better
- Sample
- 1205
- Source type
- independent-evaluator
- Evaluator
- Arena Image Edit leaderboard
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Independent evaluation. Arena public leaderboard, Image Edit category 3D Imaging & Modeling; the leaderboard score comes from anonymous side-by-side community preference votes. Its exact row uses modelDisplayName gemini-nano-banana-2.1 and links to Google documentation for Nano Banana 2.1. The row is tagged releaseType pre_release and visibly marked Preliminary. The table records the rounded public score and 1205 model votes; those votes are comparisons, not prompt samples.
- Context
- Arena marks the exact model entry preliminary. Its public table displays a +/- interval, but the checked methodology pages do not identify its confidence level, so no 95% interval is asserted here. Votes are model comparisons, not benchmark prompts; each category is a separate ballot pool.
Arena Image Edit, Multi-Image: Cartoon, Anime & Fantasy (Show measurement, test conditions, and source)
- Source value
- 1453
- Score
- 1453
- Metric
- Bradley-Terry Arena score
- Unit
- points
- Benchmark version
- Arena public leaderboard (live page)
- Category
- image-editing
- Direction
- Higher is better
- Sample
- 2640
- Source type
- independent-evaluator
- Evaluator
- Arena Image Edit leaderboard
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Independent evaluation. Arena public leaderboard, Image Edit category Cartoon, Anime & Fantasy; the leaderboard score comes from anonymous side-by-side community preference votes. Its exact row uses modelDisplayName gemini-nano-banana-2.1 and links to Google documentation for Nano Banana 2.1. The row is tagged releaseType pre_release and visibly marked Preliminary. The table records the rounded public score and 2640 model votes; those votes are comparisons, not prompt samples.
- Context
- Arena marks the exact model entry preliminary. Its public table displays a +/- interval, but the checked methodology pages do not identify its confidence level, so no 95% interval is asserted here. Votes are model comparisons, not benchmark prompts; each category is a separate ballot pool.
Arena Image Edit, Multi-Image: Overall (Show measurement, test conditions, and source)
- Source value
- 1431
- Score
- 1431
- Metric
- Bradley-Terry Arena score
- Unit
- points
- Benchmark version
- Arena public leaderboard (live page)
- Category
- image-editing
- Direction
- Higher is better
- Sample
- 10328
- Source type
- independent-evaluator
- Evaluator
- Arena Image Edit leaderboard
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Independent evaluation. Arena public leaderboard, Image Edit category Multi-Image Edit; the leaderboard score comes from anonymous side-by-side community preference votes. Its exact row uses modelDisplayName gemini-nano-banana-2.1 and links to Google documentation for Nano Banana 2.1. The row is tagged releaseType pre_release and visibly marked Preliminary. The table records the rounded public score and 10328 model votes; those votes are comparisons, not prompt samples.
- Context
- Arena marks the exact model entry preliminary. Its public table displays a +/- interval, but the checked methodology pages do not identify its confidence level, so no 95% interval is asserted here. Votes are model comparisons, not benchmark prompts; each category is a separate ballot pool.
Arena Image Edit, Multi-Image: Photorealistic & Cinematic Imagery (Show measurement, test conditions, and source)
- Source value
- 1422
- Score
- 1422
- Metric
- Bradley-Terry Arena score
- Unit
- points
- Benchmark version
- Arena public leaderboard (live page)
- Category
- image-editing
- Direction
- Higher is better
- Sample
- 6498
- Source type
- independent-evaluator
- Evaluator
- Arena Image Edit leaderboard
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Independent evaluation. Arena public leaderboard, Image Edit category Photorealistic & Cinematic Imagery; the leaderboard score comes from anonymous side-by-side community preference votes. Its exact row uses modelDisplayName gemini-nano-banana-2.1 and links to Google documentation for Nano Banana 2.1. The row is tagged releaseType pre_release and visibly marked Preliminary. The table records the rounded public score and 6498 model votes; those votes are comparisons, not prompt samples.
- Context
- Arena marks the exact model entry preliminary. Its public table displays a +/- interval, but the checked methodology pages do not identify its confidence level, so no 95% interval is asserted here. Votes are model comparisons, not benchmark prompts; each category is a separate ballot pool.
Arena Image Edit, Multi-Image: Portraits (Show measurement, test conditions, and source)
- Source value
- 1403
- Score
- 1403
- Metric
- Bradley-Terry Arena score
- Unit
- points
- Benchmark version
- Arena public leaderboard (live page)
- Category
- image-editing
- Direction
- Higher is better
- Sample
- 4885
- Source type
- independent-evaluator
- Evaluator
- Arena Image Edit leaderboard
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Independent evaluation. Arena public leaderboard, Image Edit category Portraits; the leaderboard score comes from anonymous side-by-side community preference votes. Its exact row uses modelDisplayName gemini-nano-banana-2.1 and links to Google documentation for Nano Banana 2.1. The row is tagged releaseType pre_release and visibly marked Preliminary. The table records the rounded public score and 4885 model votes; those votes are comparisons, not prompt samples.
- Context
- Arena marks the exact model entry preliminary. Its public table displays a +/- interval, but the checked methodology pages do not identify its confidence level, so no 95% interval is asserted here. Votes are model comparisons, not benchmark prompts; each category is a separate ballot pool.
Arena Image Edit, Multi-Image: Product, Branding & Commercial Design (Show measurement, test conditions, and source)
- Source value
- 1490
- Score
- 1490
- Metric
- Bradley-Terry Arena score
- Unit
- points
- Benchmark version
- Arena public leaderboard (live page)
- Category
- image-editing
- Direction
- Higher is better
- Sample
- 2389
- Source type
- independent-evaluator
- Evaluator
- Arena Image Edit leaderboard
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Independent evaluation. Arena public leaderboard, Image Edit category Product, Branding & Commercial Design; the leaderboard score comes from anonymous side-by-side community preference votes. Its exact row uses modelDisplayName gemini-nano-banana-2.1 and links to Google documentation for Nano Banana 2.1. The row is tagged releaseType pre_release and visibly marked Preliminary. The table records the rounded public score and 2389 model votes; those votes are comparisons, not prompt samples.
- Context
- Arena marks the exact model entry preliminary. Its public table displays a +/- interval, but the checked methodology pages do not identify its confidence level, so no 95% interval is asserted here. Votes are model comparisons, not benchmark prompts; each category is a separate ballot pool.
Arena Image Edit, Multi-Image: Text Rendering (Show measurement, test conditions, and source)
- Source value
- 1497
- Score
- 1497
- Metric
- Bradley-Terry Arena score
- Unit
- points
- Benchmark version
- Arena public leaderboard (live page)
- Category
- image-editing
- Direction
- Higher is better
- Sample
- 1813
- Source type
- independent-evaluator
- Evaluator
- Arena Image Edit leaderboard
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Independent evaluation. Arena public leaderboard, Image Edit category Text Rendering; the leaderboard score comes from anonymous side-by-side community preference votes. Its exact row uses modelDisplayName gemini-nano-banana-2.1 and links to Google documentation for Nano Banana 2.1. The row is tagged releaseType pre_release and visibly marked Preliminary. The table records the rounded public score and 1813 model votes; those votes are comparisons, not prompt samples.
- Context
- Arena marks the exact model entry preliminary. Its public table displays a +/- interval, but the checked methodology pages do not identify its confidence level, so no 95% interval is asserted here. Votes are model comparisons, not benchmark prompts; each category is a separate ballot pool.
Arena Image Edit, Single-Image: 3D Imaging & Modeling (Show measurement, test conditions, and source)
- Source value
- 1466
- Score
- 1466
- Metric
- Bradley-Terry Arena score
- Unit
- points
- Benchmark version
- Arena public leaderboard (live page)
- Category
- image-editing
- Direction
- Higher is better
- Sample
- 1477
- Source type
- independent-evaluator
- Evaluator
- Arena Image Edit leaderboard
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Independent evaluation. Arena public leaderboard, Image Edit category 3D Imaging & Modeling; the leaderboard score comes from anonymous side-by-side community preference votes. Its exact row uses modelDisplayName gemini-nano-banana-2.1 and links to Google documentation for Nano Banana 2.1. The row is tagged releaseType pre_release and visibly marked Preliminary. The table records the rounded public score and 1477 model votes; those votes are comparisons, not prompt samples.
- Context
- Arena marks the exact model entry preliminary. Its public table displays a +/- interval, but the checked methodology pages do not identify its confidence level, so no 95% interval is asserted here. Votes are model comparisons, not benchmark prompts; each category is a separate ballot pool.
Arena Image Edit, Single-Image: Art (Show measurement, test conditions, and source)
- Source value
- 1422
- Score
- 1422
- Metric
- Bradley-Terry Arena score
- Unit
- points
- Benchmark version
- Arena public leaderboard (live page)
- Category
- image-editing
- Direction
- Higher is better
- Sample
- 319
- Source type
- independent-evaluator
- Evaluator
- Arena Image Edit leaderboard
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Independent evaluation. Arena public leaderboard, Image Edit category Art; the leaderboard score comes from anonymous side-by-side community preference votes. Its exact row uses modelDisplayName gemini-nano-banana-2.1 and links to Google documentation for Nano Banana 2.1. The row is tagged releaseType pre_release and visibly marked Preliminary. The table records the rounded public score and 319 model votes; those votes are comparisons, not prompt samples.
- Context
- Arena marks the exact model entry preliminary. Its public table displays a +/- interval, but the checked methodology pages do not identify its confidence level, so no 95% interval is asserted here. Votes are model comparisons, not benchmark prompts; each category is a separate ballot pool.
Arena Image Edit, Single-Image: Cartoon, Anime & Fantasy (Show measurement, test conditions, and source)
- Source value
- 1462
- Score
- 1462
- Metric
- Bradley-Terry Arena score
- Unit
- points
- Benchmark version
- Arena public leaderboard (live page)
- Category
- image-editing
- Direction
- Higher is better
- Sample
- 3076
- Source type
- independent-evaluator
- Evaluator
- Arena Image Edit leaderboard
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Independent evaluation. Arena public leaderboard, Image Edit category Cartoon, Anime & Fantasy; the leaderboard score comes from anonymous side-by-side community preference votes. Its exact row uses modelDisplayName gemini-nano-banana-2.1 and links to Google documentation for Nano Banana 2.1. The row is tagged releaseType pre_release and visibly marked Preliminary. The table records the rounded public score and 3076 model votes; those votes are comparisons, not prompt samples.
- Context
- Arena marks the exact model entry preliminary. Its public table displays a +/- interval, but the checked methodology pages do not identify its confidence level, so no 95% interval is asserted here. Votes are model comparisons, not benchmark prompts; each category is a separate ballot pool.
Arena Image Edit, Single-Image: Overall (Show measurement, test conditions, and source)
- Source value
- 1428
- Score
- 1428
- Metric
- Bradley-Terry Arena score
- Unit
- points
- Benchmark version
- Arena public leaderboard (live page)
- Category
- image-editing
- Direction
- Higher is better
- Sample
- 12985
- Source type
- independent-evaluator
- Evaluator
- Arena Image Edit leaderboard
- Status
- active
- Retrieved at
- 2026-10-07
- Methodology
- Independent evaluation. Arena public leaderboard, Image Edit category Single Image Edit; the leaderboard score comes from anonymous side-by-side community preference votes. Its exact row uses modelDisplayName gemini-nano-banana-2.1 and links to Google documentation for Nano Banana 2.1. The row is tagged releaseType pre_release and visibly marked Preliminary. The table records the rounded public score and 12985 model votes; those votes are comparisons, not prompt samples.
- Context
- Arena marks the exact model entry preliminary. Its public table displays a +/- interval, but the checked methodology pages do not identify its confidence level, so no 95% interval is asserted here. Votes are model comparisons, not benchmark prompts; each category is a separate ballot pool.
Page 1 of 5