The evidence supports a platform-level release record, not a general verdict on image quality.
The record is deliberately limited to the source’s verifiable platform listing; any future performance comparison or rollout confirmation should be tracked as a separate update.
Thesis
Nano Banana 2.1 is a documented image generation and editing model on Google’s Agent Platform, with a mixed capability profile rather than an all-purpose interface.
Observed facts
Mechanism
The platform exposes image generation and multi-turn editing alongside video input, C2PA credentials, and virtual try-on, while explicitly excluding structured output, function calling, and person generation. This combination determines which workflows can be assembled directly around the model and which require separate orchestration or another model.
Why now
Google’s model documentation now lists a distinct 2.1 model ID and links to Agent Studio and Model Garden, making it actionable for platform users.
Implications
developers can evaluate a model-specific image workflow in Agent Platform
those relying on structured output, function calling, or person generation must account for the listed gaps
Evidence boundary
This documentation does not establish that 2.1 outperforms every earlier Google image model, nor does it verify availability in Gemini consumer products, AI Studio, or Flow.
Watch conditions
Check for benchmark methodology, comparative results, pricing details, and product-specific availability in Google’s own documentation before making broader claims.