LFM2-24B-A2B

This model is deprecated. Use LFM2.5-24B-A2B for improved performance.

LFM2-24B-A2B was the largest model in the LFM2 series, a Mixture-of-Experts model combining 24B total parameters with only 2B active parameters per forward pass, built for laptops and single-GPU deployments. It has been superseded by LFM2.5-24B-A2B.

HF GGUF MLX ONNX

Specifications

Property	Value
Parameters	24B (2B active)
Context Length	32K tokens
Architecture	LFM2 (MoE)

MoE Efficiency

24B quality, 2B inference cost

Laptop-Ready

Runs on laptops and single GPUs

Tool Calling

Native function calling support

Quick Start

Transformers
llama.cpp
vLLM

Install:

pip install "transformers>=5.2.0" torch accelerate

Download & Run:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "LiquidAI/LFM2-24B-A2B"
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=0.1, top_k=50, 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)

Models

Fine-tuning

Edge Inference

GPU Inference

Cloud Inference

Guides

Help

Specifications

MoE Efficiency

Laptop-Ready

Tool Calling

Quick Start

​Specifications

MoE Efficiency

Laptop-Ready

Tool Calling

​Quick Start

Specifications

Quick Start