Atlas

Models

← All models

Gemma 3n E2B

Google · 2025-06-26 · 5.4B parameters

Gemma 3n E2B is the smaller pre-trained checkpoint from Google's mobile-first Gemma 3n family, fully released on June 26, 2025 alongside its instruction-tuned variant. Its 5.44B-parameter checkpoint combines a 30-layer MatFormer text decoder with Per-Layer Embeddings, LAuReL and AltUp plus MobileNet-V5 vision and USM-derived audio encoders, accepts text, image, audio and video in a 32K-token context, and generates text; Google reports an effective memory load of about 2B parameters.

Benchmark scores

BenchmarkScore
ARC-Challenge51.7
ARC-Easy75.8
BBH (BIG-Bench Hard)44.3
BoolQ76.4
Capability85.2
DROP53.9
HellaSwag (Unspecified Scoring Protocol)72.2
Natural Questions (open-domain, closed-book)15.5
PIQA78.9
Social IQa (SIQA)48.8
TriviaQA60.8
WinoGrande66.8
Loading Atlas data…