Qwen3-Embedding-8B
A large-scale text embedding model released by Alibaba's Tongyi Qianwen team, licensed under Apache 2.0. It supports a 32K token context and 4096-dimensional vectors, and offers leading retrieval quality.
Timeline (last 90 days)
In testing on the full BEIR SciFact dataset (5,183 documents, 300 queries), Qwen3-Embedding-8B ranked first with 0.7953 nDCG@10, with EmbeddingGemma-300M close behind at 0.7861.
Based on measurements with chunks of about 250 words on an RTX A6000, all-MiniLM-L6-v2 processes 1,074.7 chunks per second, taking about 16 minutes for 1 million chunks; Qwen3-Embedding-8B processes 26.1 chunks per second, taking about 10.6 hours for 1 million chunks.
Embedding models require far less VRAM than chat models; in the table, apart from Qwen3-Embedding-8B, all models measured at bfloat16 precision peak below 2.5 GiB of VRAM, quantized Ollama files are about 1.2 GB or less, and a laptop with 8GB of memory can run them.
All Facts (3)
In testing on the full BEIR SciFact dataset (5,183 documents, 300 queries), Qwen3-Embedding-8B ranked first with 0.7953 nDCG@10, with EmbeddingGemma-300M close behind at 0.7861.
50%UnverifiedBased on measurements with chunks of about 250 words on an RTX A6000, all-MiniLM-L6-v2 processes 1,074.7 chunks per second, taking about 16 minutes for 1 million chunks; Qwen3-Embedding-8B processes 26.1 chunks per second, taking about 10.6 hours for 1 million chunks.
50%UnverifiedEmbedding models require far less VRAM than chat models; in the table, apart from Qwen3-Embedding-8B, all models measured at bfloat16 precision peak below 2.5 GiB of VRAM, quantized Ollama files are about 1.2 GB or less, and a laptop with 8GB of memory can run them.
50%