Gemma 3n

Training compute

5.2×10²³ FLOP

Parameters

7.8B

Published

May 20, 2025

Gemma 3n is an AI model developed by Google (United States), first published in May 2025. It works in the language, multimodal, speech and vision domain, on tasks such as language modeling/generation, question answering, chat and 7 more.

Training it took an estimated 5.2×10²³ FLOP of compute (estimation method: operation counting). The model has 7,850,000,000 parameters. It was trained on roughly 11T datapoints.

Access: Open weights (restricted use). Its weights are openly available. Epoch AI rates the confidence of this record as confident.

Full record

Organization: Google
Country of organization: United States
Domain: Language, Multimodal, Speech, Vision
Task: Language modeling/generation, Question answering, Chat, Speech recognition (ASR), Translation, Speech-to-text, Visual question answering, Mathematical reasoning, Code generation, Character recognition (OCR)
Training compute: 5.2×10²³ FLOP
Compute estimation method: Operation counting
Parameters: 7,850,000,000
Dataset size: 11T
Model accessibility: Open weights (restricted use)
Open weights: Yes
Epoch confidence: Confident