Model reference · open weights

embeddinggemma-qat-q8_0-unquantized

embeddinggemma-qat-q8_0-unquantized is an open-weight embedding model from Google. embeddinggemma-300m-qat-q8_0-unquantized (FP32) weighs 606 MB; the smallest configuration that runs it is RTX 3060 12 GB.

• EmbeddingGemma is a 303M parameter open embedding model developed by Google for sentence-similarity tasks.
• It generates vector representations of text to support search, retrieval, classification, and clustering applications.
• The model supports a maximum input context length of 2048 tokens and was trained on data in over 100 spoken languages.
• It is distributed under the gemma license.

Summary of the google/embeddinggemma-300m-qat-q8_0-unquantized model card, 2026-10-01

Hugging Face ↗GitHub ↗Paper ↗