The Instructor model produces text embeddings tailored to any task by providing natural language instructions, enabling classification, retrieval, clustering, and semantic similarity without model fine-tuning. It achieves state-of-the-art results across 70 diverse embedding benchmarks in science, finance, and other domains. The repository provides pre-trained checkpoints, Python utilities for encoding text, and supports use cases including custom embedding generation, similarity computation, and information retrieval pipelines.