README.md

---
library_name: transformers
license: other
license_name: lfm1.0
license_link: LICENSE
language:
- en
- ar
- zh
- fr
- de
- ja
- ko
- es
pipeline_tag: text-generation
tags:
- liquid
- lfm2
- edge
---

<center>
<div style="text-align: center;">
  <img 
    src="https://cdn-uploads.huggingface.co/production/uploads/61b8e2ba285851687028d395/2b08LKpev0DNEk6DlnWkY.png" 
    alt="Liquid AI"
    style="width: 100%; max-width: 100%; height: auto; display: inline-block; margin-bottom: 0.5em; margin-top: 0.5em;"
  />
</div>
<div style="display: flex; justify-content: center; gap: 0.5em;">
  <a href="https://playground.liquid.ai/chat">
<a href="https://playground.liquid.ai/"><strong>Try LFM</strong></a> • <a href="https://docs.liquid.ai/lfm"><strong>Documentation</strong></a> • <a href="https://leap.liquid.ai/"><strong>LEAP</strong></a></a>
</div>
</center>

# LFM2-2.6B

LFM2 is a new generation of hybrid models developed by [Liquid AI](https://www.liquid.ai/), specifically designed for edge AI and on-device deployment. It sets a new standard in terms of quality, speed, and memory efficiency. 

We're releasing the weights of four post-trained checkpoints with 350M, 700M, 1.2B, and 2.6B parameters. They provide the following key features to create AI-powered edge applications:

* **Fast training & inference** – LFM2 achieves 3x faster training compared to its previous generation. It also benefits from 2x faster decode and prefill speed on CPU compared to Qwen3.
* **Best performance** – LFM2 outperforms similarly-sized models across multiple benchmark categories, including knowledge, mathematics, instruction following, and multilingual capabilities.
* **New architecture** – LFM2 is a new hybrid Liquid model with multiplicative gates and short convolutions.
* **Flexible deployment** – LFM2 runs efficiently on CPU, GPU, and NPU hardware for flexible deployment on smartphones, laptops, or vehicles.

Find more information about LFM2 in our [blog post](https://www.liquid.ai/blog/liquid-foundation-models-v2-our-second-series-of-generative-ai-models).

## 📄 Model details

Due to their small size, **we recommend fine-tuning LFM2 models on narrow use cases** to maximize performance. 
They are particularly suited for agentic tasks, data extraction, RAG, creative writing, and multi-turn conversations. 
However, we do not recommend using them for tasks that are knowledge-intensive or require programming skills.

| Property            | [**LFM2-350M**](https://huggingface.co/LiquidAI/LFM2-350M) | [**LFM2-700M**](https://huggingface.co/LiquidAI/LFM2-700M) | [**LFM2-1.2B**](https://huggingface.co/LiquidAI/LFM2-1.2B) | [**LFM2-2.6B**](https://huggingface.co/LiquidAI/LFM2-2.6B) |
| ------------------- | ----------------------------- | ----------------------------- | ----------------------------- | ----------------------------- |
| **Parameters**      | 354,483,968                   | 742,489,344                   | 1,170,340,608                 | 2,569,272,320                 |
| **Layers**          | 16 (10 conv + 6 attn)         | 16 (10 conv + 6 attn)         | 16 (10 conv + 6 attn)         | 30 (22 conv + 8 attn)         |
| **Context length**  | 32,768 tokens                 | 32,768 tokens                 | 32,768 tokens                 | 32,768 tokens                 |
| **Vocabulary size** | 65,536                        | 65,536                        | 65,536                        | 65,536                        |
| **Precision**       | bfloat16                      | bfloat16                      | bfloat16                      | bfloat16                      |
| **Training budget** | 10 trillion tokens            | 10 trillion tokens            | 10 trillion tokens            | 10 trillion tokens            |
| **License**         | LFM Open License v1.0         | LFM Open License v1.0         | LFM Open License v1.0         | LFM Open License v1.0         

**Supported languages**: English, Arabic, Chinese, French, German, Japanese, Korean, and Spanish.

**Generation parameters**: We recommend the following parameters:
* `temperature=0.3`
* `min_p=0.15`
* `repetition_penalty=1.05`

**Reasoning**: LFM2-2.6B is the only model in this family to use dynamic hybrid reasoning (traces between `<think>` and `</think>` tokens) for complex or multilingual prompts.

**Chat template**: LFM2 uses a ChatML-like chat template as follows:

```
<|startoftext|><|im_start|>system
You are a helpful assistant trained by Liquid AI.<|im_end|>
<|im_start|>user
What is C. elegans?<|im_end|>
<|im_start|>assistant
It's a tiny nematode that lives in temperate soil environments.<|im_end|>
```

You can automatically apply it using the dedicated [`.apply_chat_template()`](https://huggingface.co/docs/transformers/en/chat_templating#applychattemplate) function from Hugging Face transformers.

**Tool use**: It consists of four main steps:
1. **Function definition**: LFM2 takes JSON function definitions as input (JSON objects between `<|tool_list_start|>` and `<|tool_list_end|>` special tokens), usually in the system prompt
2. **Function call**: LFM2 writes Pythonic function calls (a Python list between `<|tool_call_start|>` and `<|tool_call_end|>` special tokens), as the assistant answer.
3. **Function execution**: The function call is executed and the result is returned (string between `<|tool_response_start|>` and `<|tool_response_end|>` special tokens), as a "tool" role.
4. **Final answer**: LFM2 interprets the outcome of the function call to address the original user prompt in plain text.

Here is a simple example of a conversation using tool use:

```
<|startoftext|><|im_start|>system
List of tools: <|tool_list_start|>[{"name": "get_candidate_status", "description": "Retrieves the current status of a candidate in the recruitment process", "parameters": {"type": "object", "properties": {"candidate_id": {"type": "string", "description": "Unique identifier for the candidate"}}, "required": ["candidate_id"]}}]<|tool_list_end|><|im_end|>
<|im_start|>user
What is the current status of candidate ID 12345?<|im_end|>
<|im_start|>assistant
<|tool_call_start|>[get_candidate_status(candidate_id="12345")]<|tool_call_end|>Checking the current status of candidate ID 12345.<|im_end|>
<|im_start|>tool
<|tool_response_start|>[{"candidate_id": "12345", "status": "Interview Scheduled", "position": "Clinical Research Associate", "date": "2023-11-20"}]<|tool_response_end|><|im_end|>
<|im_start|>assistant
The candidate with ID 12345 is currently in the "Interview Scheduled" stage for the position of Clinical Research Associate, with an interview date set for 2023-11-20.<|im_end|>
```

You can directly pass tools as JSON schema or Python functions with `.apply_chat_template()` as shown in [this page](https://huggingface.co/docs/transformers/en/chat_extras) to automatically format the system prompt.

**Architecture**: Hybrid model with multiplicative gates and short convolutions: 10 double-gated short-range LIV convolution blocks and 6 grouped query attention (GQA) blocks.

**Pre-training mixture**: Approximately 75% English, 20% multilingual, and 5% code data sourced from the web and licensed materials.

**Training approach**:
* Very large-scale SFT on 50% downstream tasks, 50% general domains
* Custom DPO with length normalization and semi-online datasets
* Iterative model merging

## 🏃 How to run LFM2

### 1. Transformers

To run LFM2, you need to install Hugging Face [`transformers`](https://github.com/huggingface/transformers) v4.55 or a more recent version as follows:

```bash
pip install -U transformers
```

Here is an example of how to generate an answer with transformers in Python:

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model and tokenizer
model_id = "LiquidAI/LFM2-2.6B"
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    torch_dtype="bfloat16",
#    attn_implementation="flash_attention_2" <- uncomment on compatible GPU
)
tokenizer = AutoTokenizer.from_pretrained(model_id)

# Generate answer
prompt = "What is C. elegans?"
input_ids = tokenizer.apply_chat_template(
    [{"role": "user", "content": prompt}],
    add_generation_prompt=True,
    return_tensors="pt",
    tokenize=True,
).to(model.device)

output = model.generate(
    input_ids,
    do_sample=True,
    temperature=0.3,
    min_p=0.15,
    repetition_penalty=1.05,
    max_new_tokens=512,
)

print(tokenizer.decode(output[0], skip_special_tokens=False))

# <|startoftext|><|im_start|>user
# What is C. elegans?<|im_end|>
# <|im_start|>assistant
# C. elegans, also known as Caenorhabditis elegans, is a small, free-living
# nematode worm (roundworm) that belongs to the phylum Nematoda.
```

You can directly run and test the model with this [Colab notebook](https://colab.research.google.com/drive/1_q3jQ6LtyiuPzFZv7Vw8xSfPU5FwkKZY?usp=sharing).

### 2. vLLM

You need to install [`vLLM`](https://github.com/vllm-project/vllm) v0.10.2 or a more recent version as follows:

```bash
uv pip install vllm==0.10.2 --extra-index-url https://wheels.vllm.ai/0.10.2/ --torch-backend=auto
```

Here is an example of how to use it for inference:

```python
from vllm import LLM, SamplingParams

prompts = [
    "What is C. elegans?",
    "Say hi in JSON format",
    "Define AI in Spanish"
]
sampling_params = SamplingParams(temperature=0.3, min_p=0.15, repetition_penalty=1.05)

llm = LLM(model="LiquidAI/LFM2-2.6B")

outputs = llm.generate(prompts, sampling_params)

for output in outputs:
    prompt = output.prompt
    generated_text = output.outputs[0].text
    print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
```

### 3. llama.cpp

You can run LFM2 with llama.cpp using its [GGUF checkpoint](https://huggingface.co/LiquidAI/LFM2-2.6B-GGUF). Find more information in the model card.

## 🔧 How to fine-tune LFM2

We recommend fine-tuning LFM2 models on your use cases to maximize performance.

| Notebook | Description | Link |
|-------|------|------|
| SFT (Unsloth) | Supervised Fine-Tuning (SFT) notebook with a LoRA adapter using Unsloth. | <a href="https://colab.research.google.com/drive/1HROdGaPFt1tATniBcos11-doVaH7kOI3?usp=sharing"><img src="https://cdn-uploads.huggingface.co/production/uploads/61b8e2ba285851687028d395/vlOyMEjwHa_b_LXysEu2E.png" width="110" alt="Colab link"></a> |
| SFT (TRL) | Supervised Fine-Tuning (SFT) notebook with a LoRA adapter using TRL. | <a href="https://colab.research.google.com/drive/1j5Hk_SyBb2soUsuhU0eIEA9GwLNRnElF?usp=sharing"><img src="https://cdn-uploads.huggingface.co/production/uploads/61b8e2ba285851687028d395/vlOyMEjwHa_b_LXysEu2E.png" width="110" alt="Colab link"></a> |
| DPO (TRL) | Preference alignment with Direct Preference Optimization (DPO) using TRL. | <a href="https://colab.research.google.com/drive/1MQdsPxFHeZweGsNx4RH7Ia8lG8PiGE1t?usp=sharing"><img src="https://cdn-uploads.huggingface.co/production/uploads/61b8e2ba285851687028d395/vlOyMEjwHa_b_LXysEu2E.png" width="110" alt="Colab link"></a> |

## 📈 Performance

LFM2 outperforms similar-sized models across different evaluation categories. We only report scores using instruct variants and non-thinking modes for consistency.

![image/png](https://cdn-uploads.huggingface.co/production/uploads/61b8e2ba285851687028d395/xQNi_QoqAWBB1vg2XR3Rh.png)

| Model                  | MMLU  | GPQA  | IFEval | IFBench | GSM8K | MGSM  | MMMLU |
| ---------------------- | ----- | ----- | ------ | ------- | ----- | ----- | ----- |
| LFM2-2.6B              | 64.42 | 26.57 | 79.56  | 22.19   | 82.41 | 74.32 | 55.39 |
| Llama-3.2-3B-Instruct  | 60.35 | 30.6  | 71.43  | 20.78   | 75.21 | 61.68 | 47.92 |
| SmolLM3-3B             | 59.84 | 26.31 | 72.44  | 17.93   | 81.12 | 68.72 | 50.02 |
| gemma-3-4b-it          | 58.35 | 29.51 | 76.85  | 23.53   | 89.92 | 87.28 | 50.14 |
| Qwen3-4B-Instruct-2507 | 72.25 | 34.85 | 85.62  | 30.28   | 68.46 | 81.76 | 60.67 |

## 📬 Contact

If you are interested in custom solutions with edge deployment, please contact [our sales team](https://www.liquid.ai/contact).

## Citation

```
@article{liquidai2025lfm2,
 title={LFM2 Technical Report},
 author={Liquid AI},
 journal={arXiv preprint arXiv:2511.23404},
 year={2025}
}
```