16M-parameter model beats Gemini Embedding, weights open-sourced
EximiusLabs released an called Fusion Embedding 1.2B Preview. Embeddings turn text into numeric vectors so that similar meanings end up close together, which powers search, recommendations, and systems. Despite using only 16 million trainable parameters, the model reportedly outperforms Google's Gemini Embedding on benchmarks.
Having far fewer parameters means much lower to train and run. The weights are openly published on , so anyone can download and try the model directly.
Key points
- EximiusLabs released the Fusion Embedding 1.2B Preview
- Claims to beat Google's Gemini Embedding using only 16 million trainable parameters
- are openly downloadable on
- Embeddings convert text into vectors used for search, recommendations, and RAG