For more details about the training protocol of this model, please refer to the Falcon-H1 technical blogpost and Technical Report.
Currently to use this model you can either rely on Hugging Face transformers, vLLM or our llama.cpp library.
Make sure to install the latest version of transformers or vllm, eventually install these packages from source:
pip install git+https://github.com/huggingface/transformers.git
For vLLM, make sure to install vllm>=0.9.0:
pip install "vllm>=0.9.0"
Refer to the snippet below to run H1 models using ๐ค transformers:
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
model_id = "tiiuae/Falcon-H1-1B-Base"
model = AutoModelForCausalLM.from_pretrained(
model_id,
torch_dtype=torch.bfloat16,
device_map="auto"
)
# Perform text generation
For vLLM, simply start a server by executing the command below:
# pip install vllm>=0.9.0
vllm serve tiiuae/Falcon-H1-1B-Instruct --tensor-parallel-size 2 --data-parallel-size 1
llama.cpp
You can find all GGUF files compatible with llama.cpp under our official collection
Falcon-H1 series perform very well on a variety of tasks, including reasoning tasks.
| Tasks | Falcon-H1-1.5B | Qwen3-1.7B | Qwen2.5-1.5B | Gemma3-1B | Llama3.2-1B | Falcon3-1B |
|---|---|---|---|---|---|---|
| General | ||||||
| BBH | 46.57 | 43.05 | 40.55 | 30.26 | 30.72 | 35.24 |
| MMLU | 61.81 | 62.46 | 61.13 | 26.33 | 32.39 | 45.14 |
| ARC-C | 53.24 | 55.72 | 54.27 | 39.33 | 39.42 | 47.87 |
| HellaSwag | 66.76 | 67.09 | 67.86 | 62.94 | 65.73 | 62.3 |
| Winogrande | 65.59 | 66.3 | 64.56 | 62.59 | 62.75 | 61.17 |
| Math | ||||||
| GSM8k | 52.01 | 70.74 | 63.0 | 2.2 | 7.05 | 34.95 |
| MATH lvl5 | 20.39 | 16.39 | 8.84 | 1.21 | 0.98 | 3.4 |
| Science | ||||||
| GPQA | 29.11 | 29.45 | 28.36 | 24.66 | 23.57 | 27.85 |
| MMLU-Pro | 35.53 | 33.81 | 28.72 | 11.31 | 11.8 | 16.11 |
| MMLU-stem | 63.37 | 61.53 | 54.93 | 27.59 | 30.19 | 40.06 |
| Code | ||||||
| HumanEval | 50.0 | 67.68 | 35.37 | 6.71 | 18.9 | 10.37 |
| HumanEval+ | 42.68 | 60.98 | 29.27 | 5.49 | 16.46 | 9.15 |
| MBPP | 65.08 | 67.72 | 60.05 | 12.7 | 35.98 | 12.43 |
| MBPP+ | 55.03 | 58.99 | 49.47 | 9.52 | 29.89 | 9.52 |
You can check more in detail on our our release blogpost, detailed benchmarks.
If the Falcon-H1 family of models were helpful to your work, feel free to give us a cite.
@article{falconh1,
title={Falcon-H1: A Family of Hybrid-Head Language Models Redefining Efficiency and Performance},
author={Jingwei Zuo and Maksim Velikanov and Ilyas Chahed and Younes Belkada and Dhia Eddine Rhayem and Guillaume Kunsch and Hakim Hacid and Hamza Yous and Brahim Farhat and Ibrahim Khadraoui and Mugariya Farooq and Giulia Campesan and Ruxandra Cojocaru and Yasser Djilali and Shi Hu and Iheb Chaabane and Puneesh Khanna and Mohamed El Amine Seddik and Ngoc Dung Huynh and Phuc Le Khac and Leen AlQadi and Billel Mokeddem and Mohamed Chami and Abdalgader Abubaker and Mikhail Lubinets and Kacper Piskorski and Slim Frikha},
journal = {arXiv preprint arXiv:2507.22448},
year={2025}
}