Try LFM • Docs • LEAP • Discord

LFM2.5-2.6B-MLX

LFM2.5 is a new family of hybrid models designed for on-device deployment. It builds on the LFM2 architecture with extended pre-training and reinforcement learning.

Find more details in the original model card: https://huggingface.co/LiquidAI/LFM2.5-2.6B

Precisions

Folder	Precision	Group Size	Size
bf16/	bf16	-	5.02 GB
8bit/	8-bit	64	2.67 GB
6bit/	6-bit	64	2.04 GB
5bit/	5-bit	64	1.76 GB
4bit/	4-bit	64	1.47 GB
mxfp8/	MXFP8	32	2.59 GB
mxfp4/	MXFP4	32	1.46 GB
nvfp4/	NVFP4	16	1.53 GB

Use with mlx

mlx_lm.load does not resolve subfolders of a HuggingFace repo directly (ml-explore/mlx-lm#403), so download the precision you want first:

pip install mlx-lm

from huggingface_hub import snapshot_download
from mlx_lm import load, generate
from mlx_lm.sample_utils import make_sampler

path = snapshot_download("LiquidAI/LFM2.5-2.6B-MLX", allow_patterns=["4bit/*"])
model, tokenizer = load(f"{path}/4bit")

response = generate(
    model,
    tokenizer,
    prompt="The capital of France is",
    max_tokens=100,
    sampler=make_sampler(temp=0.7),
    verbose=True,
)