The default local RAG embedding model for 2026. 100 languages, 8K context, and three retrieval modes (dense, sparse, multi-vector) in one model. MIT license. Used in production by thousands of enterprise RAG pipelines.
This model requires a specialized High-VRAM environment. Ensure you have the latest CUDA Drivers or Metal Framework installed.
Minimum VRAM: 3GB VRAM Recommended
The BGE-M3 model by BAAI is a 568M parameter architecture optimized for embedding tasks. It requires approximately 0.5GB of VRAM to comfortably run locally using a FP16 quantization. Extending the context window up to 8,192 tokens will dynamically allocate further VRAM, meaning high-bandwidth memory hardware is strictly advised.
For running BGE-M3 at maximum tokens-per-second, we recommend using LM Studio or Ollama with a GGUF quantization (Q4_K_M or Q6_K). If you are multi-GPU, use vLLM to distribute the layers across your VRAM pool for optimal throughput.
You will need a GPU with at least 3GB of VRAM to run the FP16 quantized version smoothly with a moderate context window.
The simplest method is utilizing Ollama by executing 'ollama run bge-m3' directly in your command line. Alternatively, you can search for the model via LM Studio's interface.