Google unveils Gemini Nano Banana 2.1 and EmbeddingGemma 2 AI models

Google introduced Nano Banana 2.1, an upgraded image generation and conversational editing model with 1K–4K output and support for up to 14 reference images, alongside EmbeddingGemma 2, a 740M-parameter on-device multimodal embedding model with an 8K-token context window. Nano Banana 2.1 is rolling out in the Gemini app, AI Mode in Search and Google AI Studio, while EmbeddingGemma 2 is available to developers via Hugging Face and Kaggle.

Digit

Open source

Google unveils Gemini Nano Banana 2.1 and EmbeddingGemma 2 AI models

Nano Banana 2.1 supports 1K, 2K and 4K image generation
It keeps up to 4 characters and 10 objects consistent, accepting up to 14 references
EmbeddingGemma 2 has 740M parameters, Gemma 4 architecture and 8K-token context
EmbeddingGemma 2 maps text, images, audio and video into a unified embedding space

#Google#Gemini#EmbeddingGemma

Google unveils Gemini Nano Banana 2.1 and EmbeddingGemma 2 AI models

Google launches Nano Banana 2.1 with better editing and subject consistency

Google unveils Gemini 4 Argon frontier model with 1M output tokens

Google DeepMind launches EmbeddingGemma 2, a 740M-parameter multimodal embedding model

Google Research unveils AI video co-director with 4 agentic frameworks