Show HN: Matching Jev on BANKING77 at a thousandth of the cost
Category: ai-ml
Tags: embeddings, intent-classification, benchmark, vector-search, nlp
Score: 5.0/10 (Innovation: 5, Technical: 5, Documentation: 3, Utility: 5)
The Tuatara Vector Model is an embedding model developed at Lawrence Berkeley National Lab that achieves near-parity with Jev's model on the BANKING77 intent classification benchmark (91.79% vs 92.40%) at a fraction of the cost. It offers a cost-effective alternative for vector-based intent classification, though it is presented primarily as a benchmark result rather than a full open-source package.
Target audience: data scientists, ML engineers, NLP practitioners
View on Hacker News