---
library_name: transformers
license: other
license_name: lfm1.0
license_link: LICENSE
language:
- en
- ar
- zh
- fr
- de
- ja
- ko
- es
- pt
pipeline_tag: text-generation
tags:
- liquid
- lfm2.5
- edge
base_model: LiquidAI/LFM2.5-350M-Base
---
# LFM2.5-350M
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.
- **Best-in-class performance**: A 350M model rivaling much larger models, bringing high-quality AI to your pocket.
- **Fast edge inference**: 313 tok/s decode on AMD CPU, 188 tok/s on Snapdragon Gen4. Runs under 1GB of memory with day-one support for llama.cpp, MLX, and vLLM.
- **Scaled training**: Extended pre-training from 10T to 28T tokens and large-scale multi-stage reinforcement learning.
Find more information about LFM2.5-350M in our [blog post](https://www.liquid.ai/blog/lfm2-5-350m-no-size-left-behind).
> [!NOTE]
> 💻 **Demo**: https://huggingface.co/spaces/webml-community/lfm2.5-webgpu-summarizer
## 🗒️ Model Details
| Model | Parameters | Description |
|-------|------------|-------------|
| [LFM2.5-350M-Base](https://huggingface.co/LiquidAI/LFM2.5-350M-Base) | 350M | Pre-trained base model for fine-tuning |
| [**LFM2.5-350M**](https://huggingface.co/LiquidAI/LFM2.5-350M) | 350M | General-purpose instruction-tuned model |
LFM2.5-350M is a general-purpose text-only model with the following features:
- **Number of parameters**: 350M
- **Number of layers**: 16 (10 double-gated LIV convolution blocks + 6 GQA blocks)
- **Training budget**: 28T tokens
- **Context length**: 32,768 tokens
- **Vocabulary size**: 65,536
- **Knowledge cutoff**: Mid-2024
- **Languages**: English, Arabic, Chinese, French, German, Japanese, Korean, Portuguese, Spanish
- **Generation parameters**:
- `temperature: 0.1`
- `top_k: 50`
- `repetition_penalty: 1.05`
| Model | Description |
|-------|-------------|
| [**LFM2.5-350M**](https://huggingface.co/LiquidAI/LFM2.5-350M) | Original model checkpoint in native format. Best for fine-tuning or inference with Transformers and vLLM. |
| [LFM2.5-350M-GGUF](https://huggingface.co/LiquidAI/LFM2.5-350M-GGUF) | Quantized format for llama.cpp and compatible tools. Optimized for CPU inference and local deployment with reduced memory usage. |
| [LFM2.5-350M-ONNX](https://huggingface.co/LiquidAI/LFM2.5-350M-ONNX) | ONNX Runtime format for cross-platform deployment. Enables hardware-accelerated inference across diverse environments (cloud, edge, mobile). |
| [LFM2.5-350M-MLX](https://huggingface.co/LiquidAI/LFM2.5-350M-MLX-8bit) | MLX format for Apple Silicon. Optimized for fast inference on Mac devices using the MLX framework. |
| [LFM2.5-350M-OpenVINO](https://huggingface.co/OpenVINO/LFM2.5-350M-int8-ov) | OpenVINO format for Intel hardware acceleration. Optimized for efficient inference on Intel CPUs, GPUs, and NPUs. |
We recommend using it for data extraction, structured outputs, and tool use. It is not recommended for knowledge-intensive tasks and programming.
### Chat Template
LFM2.5 uses a ChatML-like format. See the [Chat Template documentation](https://docs.liquid.ai/lfm/key-concepts/chat-template) for details. Example:
```
<|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
```
You can use [`tokenizer.apply_chat_template()`](https://huggingface.co/docs/transformers/en/chat_templating#using-applychattemplate) to format your messages automatically.
### Tool Use
LFM2.5 supports function calling as follows:
1. **Function definition**: We recommend providing the list of tools as a JSON object in the system prompt. You can also use the [`tokenizer.apply_chat_template()`](https://huggingface.co/docs/transformers/en/chat_extras#passing-tools) function with tools.
2. **Function call**: By default, LFM2.5 writes Pythonic function calls (a Python list between `<|tool_call_start|>` and `<|tool_call_end|>` special tokens), as the assistant answer. You can override this behavior by asking the model to output JSON function calls in the system prompt.
3. **Function execution**: The function call is executed, and the result is returned as a "tool" role.
4. **Final answer**: LFM2 interprets the outcome of the function call to address the original user prompt in plain text.
See the [Tool Use documentation](https://docs.liquid.ai/lfm/key-concepts/tool-use) for the full guide. Example:
```
<|startoftext|><|im_start|>system
List of tools: [{"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"]}}]<|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
[{"candidate_id": "12345", "status": "Interview Scheduled", "position": "Clinical Research Associate", "date": "2023-11-20"}]<|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|>
```
## 🏃 Inference
LFM2.5 is supported by many inference frameworks. See the [Inference documentation](https://docs.liquid.ai/lfm/inference/transformers) for the full list.
| Name | Description | Docs | Notebook |
|------|-------------|------|:--------:|
| [Transformers](https://github.com/huggingface/transformers) | Simple inference with direct access to model internals. | Link | 
|
| [vLLM](https://github.com/vllm-project/vllm) | High-throughput production deployments with GPU. | Link | |
| [llama.cpp](https://github.com/ggml-org/llama.cpp) | Cross-platform inference with CPU offloading. | Link | |
| [MLX](https://github.com/ml-explore/mlx) | Apple's machine learning framework optimized for Apple Silicon. | Link | — |
| [LM Studio](https://lmstudio.ai/) | Desktop application for running LLMs locally. | Link | — |
| [OpenVINO](https://www.intel.com/content/www/us/en/developer/tools/openvino-toolkit/overview.html) | Intel's toolkit for optimized inference on CPUs, GPUs, and NPUs. | Link | — |
Here's a quick start example with Transformers:
```python
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
model_id = "LiquidAI/LFM2.5-350M"
model = AutoModelForCausalLM.from_pretrained(
model_id,
device_map="auto",
dtype="bfloat16",
# attn_implementation="flash_attention_2" <- uncomment on compatible GPU
)
tokenizer = AutoTokenizer.from_pretrained(model_id)
streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
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=0.1,
top_k=50,
repetition_penalty=1.05,
max_new_tokens=512,
streamer=streamer,
)
```
## 🔧 Fine-Tuning
We recommend fine-tuning LFM2.5 for your specific use case to achieve the best results.
| Name | Description | Docs | Notebook |
|------|-------------|------|----------|
| CPT ([Unsloth](https://github.com/unslothai/unsloth)) | Continued Pre-Training using Unsloth for text completion. | Link | |
| CPT ([Unsloth](https://github.com/unslothai/unsloth)) | Continued Pre-Training using Unsloth for translation. | Link | |
| SFT ([Unsloth](https://github.com/unslothai/unsloth)) | Supervised Fine-Tuning with LoRA using Unsloth. | Link | |
| SFT ([TRL](https://github.com/huggingface/trl)) | Supervised Fine-Tuning with LoRA using TRL. | Link | |
| DPO ([TRL](https://github.com/huggingface/trl)) | Direct Preference Optimization with LoRA using TRL. | Link | |
| GRPO ([Unsloth](https://github.com/unslothai/unsloth)) | GRPO with LoRA using Unsloth. | Link | |
| GRPO ([TRL](https://github.com/huggingface/trl)) | GRPO with LoRA using TRL. | Link | |
## 📊 Performance
### Benchmarks
| Model | GPQA Diamond | MMLU-Pro | IFEval | IFBench | Multi-IF |
|---|---|---|---|---|---|
| LFM2.5-350M | 30.64 | 20.01 | 76.96 | 40.69 | 44.92 |
| LFM2-350M | 27.58 | 19.29 | 64.96 | 18.20 | 32.92 |
| Granite 4.0-H-350M | 22.32 | 13.14 | 61.27 | 17.22 | 28.70 |
| Granite 4.0-350M | 25.91 | 12.84 | 53.48 | 15.98 | 24.21 |
| Qwen3.5-0.8B (Instruct) | 27.41 | 37.42 | 59.94 | 22.87 | 41.68 |
| Qwen3.5-0.8B (Thinking) | 19.29 | -* | 32.93 | 22.00 | 26.44 |
| Gemma 3 1B IT | 23.89 | 14.04 | 63.49 | 20.33 | 44.25 |
| Model | CaseReportBench | BFCLv3 | BFCLv4 | τ²-Bench Telecom | τ²-Bench Retail |
|---|---|---|---|---|---|
| LFM2.5-350M | 32.45 | 44.11 | 21.86 | 18.86 | 17.84 |
| LFM2-350M | 11.67 | 22.95 | 12.29 | 10.82 | 5.56 |
| Granite 4.0-H-350M | 12.44 | 43.07 | 13.28 | 13.74 | 6.14 |
| Granite 4.0-350M | 0.84 | 39.58 | 13.73 | 2.92 | 6.14 |
| Qwen3.5-0.8B (Instruct) | 13.83 | 35.08 | 18.70 | 12.57 | 6.14 |
| Qwen3.5-0.8B (Thinking) | 0.39 | 39.64 | 25.39 | 14.33 | 7.02 |
| Gemma 3 1B IT | 2.28 | 16.61 | 7.17 | 9.36 | 6.43 |
*Evaluation could not be completed due to doom looping.
### CPU Inference
### GPU Inference
## 📬 Contact
- Got questions or want to connect? [Join our Discord community](https://discord.com/invite/liquid-ai)
- If you are interested in custom solutions with edge deployment, please contact [our sales team](https://www.liquid.ai/contact).
## Citation
```bibtex
@article{liquidAI2026350M,
author = {Liquid AI},
title = {LFM2.5-350M: No Size Left Behind},
journal = {Liquid AI Blog},
year = {2026},
note = {www.liquid.ai/blog/lfm2-5-350m-no-size-left-behind},
}
```
```bibtex
@article{liquidai2025lfm2,
title={LFM2 Technical Report},
author={Liquid AI},
journal={arXiv preprint arXiv:2511.23404},
year={2025}
}
```