Google ships Nano Banana 2.1: multi-character consistency jumps to 1106, Infographic Factuality nearly triples
机器之心 · wechat · 2026-10-07
Google DeepMind released Nano Banana 2.1 (model ID: gemini-nano-banana-2.1), built on Gemini 3.6 Flash and positioned as an efficient image generation and conversational editing model, now stable in the Gemini API and rolled out across Gemini App, AI Studio, Google Ads, Flow and Stitch.
Key upgrades:
- Design quality: handles full layouts with fonts, hierarchy and whitespace; Infographic Design score 1048 vs 961 (Nano Banana 2) and 912 (Pro)
- Mask-based editing: 1049 vs 965/927; General Editing up from 938/939 to 1026
- Subject consistency: single-character 981→1028, multi-character 978→1106, beating even Nano Banana Pro
- Natural imagery: improved realism at 1K/2K/4K; new wide aspect ratios with tile-artifact fixes; up to 14 reference images
- Infographics: grounds on Gemini world knowledge plus Google Search; Infographic Factuality 0.179→0.521
Remaining limits: blurry small text, weak long-paragraph rendering, incomplete edits, occasional spatial errors, and gaps in 3D reasoning and factuality. Read as a workflow-focused 0.1 upgrade rather than a big capability jump.
More from Multimodal
- Spira Maxima ties open-source release to 2M views or 5K followers — nikola_mr64990 · 2026-10-07
- Google ships Nano Banana 2.1, jumps 80 points on LMArena's text-to-image board — danielrock · 2026-10-07
- DRAMA 1.0 launches: edit only the expression layer, swap 10 emotions in existing footage — nikola_mr64990 · 2026-10-07
- video-shotcraft hits 10.5k stars: cinematic product videos via Claude Code + Remotion — tom_doerr · 2026-10-07
- Third-party test: Google's Nano Banana 2.1 beats ChatGPT Images, 5x faster and cheaper — prajdabre · 2026-10-07
- Single Image to Full 3D Scene: Adaptive Chunking Extends Object Generators to Outdoor Rome — Jiraphon Yenphraphai · 2026-10-07