Nano Banana 2.1
Google's latest Nano Banana image model (Gemini API code gemini-nano-banana-2.1), released on October 6, 2026 as an update to Nano Banana 2 with better quality, prompt adherence and text rendering. It is in the tied second group on LMArena and fourth on Artificial Analysis' image arena.
GoogleImage modelReleased October 2026Data checked October 18, 2026
This model is not in the current people's rankings: it is either new and still collecting votes, or ranked below the top 12. See the AI leaderboard for the models people rate highest.
Strengths and weaknesses
- Tied for second on LMArena text-to-image and image editing
- Fourth on the Artificial Analysis text-to-image arena
- Cheap: $0.0336 per 1K image on the Gemini API
- Very new, with fewer LMArena votes than older models
- Well behind GPT-Image-2.5 Sunburst on LMArena
Best for
- Everyday image creation in Gemini
- Infographics and images with text
- Keeping characters consistent across edits
Every score we have
Each bar compares this model with the best result any model on this page has on the same test.
- LMArena text to image1,327.8Arena score 1327.8 (95% CI 1319.0 to 1336.5); 5,310 votes; entry "gemini-nano-banana-2.1"; position 6 of 82 on the full board
- LMArena image editing1,428.1Arena score 1428.1 (95% CI 1422.3 to 1433.9); 12,985 votes; entry "gemini-nano-banana-2.1"; position 6 of 59 on the full board
- Artificial Analysis image arena1,174.3Elo 1174.3 (95% CI 1162.3 to 1186.3); 5,961 appearances; position 4 of 104 on the Artificial Analysis text-to-image arena
- Price per 1,000 images (estimate)$34$33.60 per 1,000 images (Artificial Analysis estimate)
Compare with
- GPT-Image-2.5 SunburstOpenAI's most capable image generation and editing model, released as a September 8, 2026 snapshot alongside the faster GPT-Image-2.5 Flare.
- Grok Imagine Image 2.0SpaceXSI's (formerly xAI) second-generation Grok Imagine image model for generating and editing images from text or image prompts.
- MAI-Image-2.6Microsoft AI's image model, launched on August 10, 2026 and offered in public preview on Microsoft Foundry.
- Muse ImageMeta's agentic image model, which plans and checks its work before rendering and can ground images in search results.