Mid-range 4B multilingual embedding workhorse, quality vs. cost sweet spot for the F2LLM family.
Access model weights, configuration files, and documentation.
See which devices can run this model and at what quality level.
The 4B mid-range model in the F2LLM-v2 family from CodeFuse-AI, fine-tuned from Qwen3-4B-Base via the family's open two-stage contrastive + instruction-tuned recipe on 60M multilingual samples. Successor to F2LLM-v1-4B, which previously ranked #2 in its size tier and #7 overall on MTEB English.
View on Hugging Face