← All models

Gemma 3n E4B

Google · 2025-06-26 · 7.8B parameters

Gemma 3n E4B is the larger pre-trained checkpoint from Google's mobile-first Gemma 3n family, fully released on June 26, 2025 alongside its instruction-tuned variant. Its MatFormer architecture nests an E2B submodel, and Per-Layer Embeddings let about 4B core transformer parameters reside in accelerator memory although the checkpoint holds about 7.85B parameters; it accepts text, image, audio and video in a 32K-token context and generates text.

Benchmark scores