Google just launched Nano Banana 2.1, its newest image generation and editing model built on Gemini 3.6 Flash. This update is now rolling out sequentially across the Gemini app, Search’s AI Mode, Google Ads, and developer platforms like Google AI Studio, Flow, Stitch, and the Gemini Enterprise Agent Platform.
Sharper 4K edits, subject consistency, and search grounding
Nano Banana 2.1 tackles some of the most annoying quirks in AI image generation. Built with a context window scaling up to 1 million tokens, it handles resolutions from 1K and 2K up to crisp 4K across aspect ratios ranging from 1:1 all the way out to ultra-wide 8:1. It also accepts direct inline data uploads up to 7MB and Cloud Storage files up to 30MB, consuming roughly 1,120 tokens per input image.
For multi-step editing, it processes up to 14 reference images at once while tracking up to four characters and ten objects. That keeps a person or object looking like itself across multiple edits instead of morphing between prompts. Developers can also adjust the inference “thinking” level and toggle grounding via Google Search and Google Image Search to verify real-world details before rendering.
Massive API price cuts for developers
Beyond technical upgrades, the biggest news for builders is the price drop. Google halved standard paid-tier API rates to $30.00 per million image tokens (down from $60.00). Meanwhile, text inputs sit at $1.50 per million tokens and text/thinking outputs run $7.50 per million tokens. For image output, that translates to $0.0336 for a 1K image (down from $0.067), $0.0504 for 2K, and $0.0756 for 4K (down from $0.151).
Batch processing jobs get another 50% discount across the board, bringing 1K generations down to just $0.0168 per image. Lastly, the model incorporates C2PA Content Credentials for digital origin tracking, though enterprise platform documentation notes that generating real people remains unavailable on that specific platform tier.
The post Google Nano Banana 2.1 Brings Sharper Editing and 50% Cost Cuts appeared first on Android Headlines.
