LFM2-350M

This model is deprecated. Use LFM2.5-350M for improved performance in the same compact model class.

Property	Value
Parameters	350M
Context Length	32K tokens
Architecture	LFM2 (Dense)

Property

Value

Parameters

350M

Context Length

32K tokens

Architecture

LFM2 (Dense)

Quick Start

Transformers
llama.cpp
vLLM
SGLang

Install:

pip install "transformers>=5.2.0" torch accelerate

Download & Run:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "LiquidAI/LFM2-350M"
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    dtype="bfloat16",
)
tokenizer = AutoTokenizer.from_pretrained(model_id)

inputs = tokenizer.apply_chat_template(
    [{"role": "user", "content": "What is machine learning?"}],
    add_generation_prompt=True,
    return_tensors="pt",
    tokenize=True,
    return_dict=True,
).to(model.device)

output = model.generate(**inputs, do_sample=True, temperature=0.3, min_p=0.15, repetition_penalty=1.05, max_new_tokens=512)
input_length = inputs["input_ids"].shape[1]
response = tokenizer.decode(output[0][input_length:], skip_special_tokens=True)
print(response)

Install:

brew install llama.cpp

Run:

llama-cli -hf LiquidAI/LFM2-350M-GGUF \
    --temp 0.3 --min-p 0.15 --repeat-penalty 1.05

The -hf flag downloads the model directly from Hugging Face. For other installation methods and advanced usage, see the llama.cpp guide.

Models

Fine-tuning

Edge Inference

GPU Inference

Cloud Inference

Guides

Help

Specifications

Ultra-Light

Low Latency

Edge-Ready

Quick Start

​Specifications

Ultra-Light

Low Latency

Edge-Ready

​Quick Start

Specifications

Quick Start