Papers with Code Search Engine Uses Qwen 0.6B and pgvector
August 25, 2026
Papers with Code implements hybrid search using the Alibaba Qwen 3-Embedding-0.6B model and PostgreSQL with pgvector. Embeddings are computed on NVIDIA L4 GPUs via Hugging Face Jobs and served through Inference Endpoints.
HOW THIS AFFECTS YOU
●
builderYou can replicate this architecture using pgvector and small embedding models to optimize search latency and cost.
●
researcherThis demonstrates an efficient deployment pattern for large-scale academic literature retrieval.