F2LLM

F2LLMs (Foundation-to-Feature Large Language Models) are foundation models directly finetuned on 6 million high-quality query-document pairs, striking a strong balance between model size, training cost, and embedding performance:

On the MTEB leaderboard, F2LLM-4B ranks 2nd among models of ~4B size, and 7th overall, while F2LLM-1.7B ranks 1st among models of 1B-2B size.

F2LLMs are fully open. Model checkpoints are available at:

Training data is available at F2LLM data.

Train

In this repo we provide a streamlined and efficient script for training embedding models. To reproduce the training of F2LLMs, please:

Setup environment following requirements.txt. We note that transformers>=4.51.0 is required for training Qwen3 models.
Download data and backbone models from Hugging Face (we use Qwen3 models).
Run tokenize_data_qwen.py to tokenize the downloaded data
Modify model path, data path, and other arguments in configs/config.json.
Start training with accelerate launch --config_file configs/accelerate_config.yaml run.py --config configs/config.json.

Note: we recommend setting num_processes to 1 in configs/accelerate_config.yaml and launch the training code once to generate cache for training data before starting the actual training.

For multi-node training, run on the main node:

accelerate launch --config_file configs/accelerate_config.yaml --num_machines N_NODE --num_processes N_PROCESSES --machine_rank 0 --main_process_ip MASTER_IP --main_process_port MASTER_PORT run.py --config configs/config.json

where N_NODE is the number of machines; N_PROCESSES is N_NODE*8; MASTER_IP is the IP address of your master node, and MASTER_PORT is a port available on your machine (e.g. 6379).

On worker nodes, also run the above commmand but modify machine_rank accordingly.

Citation

If you use the F2LLM models, data, or code, please cite the following technical report.

@article{2025F2LLM,
  title={F2LLM Technical Report: Matching SOTA Embedding Performance with 6 Million Open-Source Data},
  author={Ziyin Zhang and Zihan Liao and Hang Yu and Peng Di and Rui Wang},
  journal      = {CoRR},
  volume       = {abs/2510.02294},
  year         = {2025},
  url          = {https://doi.org/10.48550/arXiv.2510.02294},
  doi          = {10.48550/ARXIV.2510.02294},
  eprinttype    = {arXiv},
  eprint       = {2510.02294}
}