Nano Banana 2.1

Nano Banana 2.1 (gemini-nano-banana-2.1), generally available October 6, 2026, is Google's current default image model across the Gemini API, AI Studio, the Gemini app, AI Mode in Search, Ads, Flow, and Stitch. Google positions it as better than Nano Banana 2 across the board — visual quality, prompt adherence, character consistency, text rendering — at half the per-image price.

It generates and edits at 1K, 2K, and 4K, takes up to 14 reference images (plus video inputs), supports Google Search grounding, multi-turn editing, virtual try-on, and person generation, and marks every image with SynthID. The context window is 131,072 tokens; note that seed, temperature, topP, topK, and logprobs are unsupported and error if set. Pricing is $1.50 per million input tokens and $30 per million image-output tokens — about $0.0336 per 1K image, halved again on the Batch API.

Nano Banana 2 is deprecated and shuts down October 29, 2026. On Arena's October 6 boards, 2.1 scored 1328 in text-to-image (#5) and 1428 in image edit — ahead of Nano Banana Pro (1248/1390) at roughly a quarter of the price, behind GPT Image 2.5 Sunburst (1425/1524). Google has published no side-by-side benchmarks for its 'better across the board' claim, so test your own prompts before migrating.

Independent (Arena, Oct 6): text-to-image 1328, image edit 1428 — ahead of Nano Banana Pro (1248/1390), behind GPT Image 2.5 Sunburst (1425/1524). Provider: 'outperforms previous models across the board' with no published measurements.