Instructions to use LiquidAI/LFM2.5-230M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use LiquidAI/LFM2.5-230M with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="LiquidAI/LFM2.5-230M")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("LiquidAI/LFM2.5-230M")
model = AutoModelForCausalLM.from_pretrained("LiquidAI/LFM2.5-230M")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use LiquidAI/LFM2.5-230M with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "LiquidAI/LFM2.5-230M"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "LiquidAI/LFM2.5-230M",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/LiquidAI/LFM2.5-230M

SGLang

How to use LiquidAI/LFM2.5-230M with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "LiquidAI/LFM2.5-230M" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "LiquidAI/LFM2.5-230M",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "LiquidAI/LFM2.5-230M" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "LiquidAI/LFM2.5-230M",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use LiquidAI/LFM2.5-230M with Docker Model Runner:
```
docker model run hf.co/LiquidAI/LFM2.5-230M
```

LFM2.5-230M beats LFM2.5-350M on GSM8K?

#10

by coder543 - opened 1 day ago

Discussion

coder543

1 day ago

•

edited 1 day ago

I converted both models to bf16 GGUFs, and used llama-eval locally to run gsm8k in the regex eval mode against them.

I saw 24.6% for LFM2.5-350M and 29.7% for LFM2.5-230M. Strangely, the 230M model is winning by a significant, repeatable margin.

The benchmarks make it seem like the 350M should be better at everything, so I find this result confusing? Maybe I did something wrong?

mlabonne

Liquid AI org about 12 hours ago

•

edited about 12 hours ago

Hey, I ran GSM8K in an avg@10 setting with temperature=0.6: LFM2.5-230M gets 29.08% and LFM2.5-350M gets 28.36%. Both models are very close across all math evals.

I'd be slightly suspicious about the score provided by llama-eval. Maybe something related to parsing?

coder543

about 10 hours ago

Yeah, I’m not sure. I was impressed at how much better LFM2.5-230M was compared to my previous experience testing out Gemma 3 270M, so then I thought maybe the 350M would be even better.

I think these tiny models are starting to get useful, so I can only imagine where they’ll be at next year!

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment