Unverified50% confidenceFactExact time
EmbeddingGemma-300M is a small open-source embedding model released by Google with about 300 million parameters, supporting multiple languages but with training data dominated by high-resource languages.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSmall language models typically have parameter counts ranging from hundreds of millions to tens of billions, such as Microsoft's Phi-3 (3.8B), Google's Gemma 2B, and Meta's Llama 3.2 (1B/3B).64% similarUnverifiedIn another internal case, a group of Google agents helped free more than 300 TB of memory by analyzing performance information.61% similarVerifiedThe concept behind Embedding technology originates from the Word2Vec model proposed by Google in 2013.60% similarPartially VerifiedEmbeddingGemma 2 is a multimodal embedding model built by Google on the Gemma 4 architecture, capable of encoding text, images, audio, and video into a shared 768-dimensional vector space.60% similarUnverifiedGoogle has released EmbeddingGemma 2, an embedding model with a total of 740M parameters, which is Google's first natively multimodal on-device open-source embedding model.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1153276API
curl https://kongchang.com/api/v1/knowledge/claims/1153276MCP
get_claim(id=1153276)