A compact, focused specialist that converts text into dense vector representations. At 0.6 billion parameters, it stays lean and fast, trading raw capacity for efficiency. It does one thing — embedding — and does it without the overhead of a generative model.