EmbeddingGemma 300M
A lightweight, state-of-the-art open embedding model.
A roughly 300-million-parameter open text-embedding model in the Gemma ecosystem, intended to turn text into embedding vectors for retrieval, semantic similarity, clustering, and related language tasks, including on-device use.
Pricing
Third-party hosted inference price (DeepInfra), not a Google first-party API price. Embedding response output is listed at USD 0 per 1M tokens by the route listing.
Google's model card identifies EmbeddingGemma as Google DeepMind's 300M open text-embedding model. Google does not publish a token-priced hosted API rate for this exact model; DeepInfra's public hosted-model page lists USD 0.002 per 1M tokens. Opper's route listing for the same DeepInfra model ID explicitly lists USD 0.0020 per 1M input tokens and USD 0 per 1M output tokens.