Nomic releases text embedding weights, code and training data

Release and source context

Nomic released nomic-embed-text-v1 with model weights and training code under Apache-2.0, alongside training data and a technical report. The embedding model supported an 8,192-token context for applications such as retrieval; the paper’s later submission date is separate from the model release. [1]

Nomic stated Apache 2.0 licensing and an 8,192-token context window. Benchmark comparisons came from the developer, and the paper's February 2 arXiv submission is separate from the February 1 release. [1]

The Decoder published its contemporary report on February 1, 2024. [2]