Google's instruction-tuned open-weights multimodal model, the largest single-accelerator-deployable model in the Gemma 3 family.

Benchmarks

What its maker published, and what anyone else measured. Every figure links the document it came from.

benchmark	published	measured
MMLU-Pro	67.5	—
GPQA	42.4	—

Lineage

No recorded lineage — root or standalone model.

view in full graph →

Metadata

org	Google
released	2025-03-12
params	27B
license	gemma
hf id	google/gemma-3-27b-it