Jina Embeddings v5 Text Nano by Jina AI — 768D

239M-parameter lightweight multilingual text embedding model scoring 71.0 on MTEB English v2 and 65.5 on MMTEB. Matches or exceeds all other sub-500M embedding models including KaLM-mini-v2.5 and Gemma-300M. Built on EuroBERT-210M with distillation from Qwen3-Embedding-4B. Supports 32K tokens, 768-dim with Matryoshka truncation. Ideal for resource-constrained deployments.