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
```

Please release LFM2.5-125M

by Tralalabs - opened 1 day ago

Discussion

Tralalabs

1 day ago

Please, release it. Would like to see smaller models in the LFM family. Say "Will do. SOON!"

FlameF0X

1 day ago

mmmm 🤤 lfm2.5-50m-thinking-max /j

Tralalabs

1 day ago

@mlabonne Please finetune qwen3.5 9b to make it frakenstein (finetuned name: QWENFRAKEN-9B) on WithinUsAI/gemini_3.5_flash_distilled_25k + WithinUsAI/GeminiPro3.2_max_distill_god_seed_25k + WithinUsAI/claude_fable_5_distillation_merged_cleaned_25k + WithinUsAI/claude_mythos_distilled_25k + Glint-Research/Fable-5-traces + armand0e/claude-fable-5-claude-code + lazarus19/Vibe-Coding-Instruct + angrygiraffe/claude-opus-4.6-4.7-reasoning-8.7k + lordx64/agentic-distill-fable-5-sft + Crownelius/Complete-FABLE.5-traces-2M + AletheiaResearch/GLM-5.2-Agent + PawanKrd/claude-fable-5-code + WithinUsAI/GPT_5.5_Distilled + HelioAI/Fable-5-Distill-Reasoning-462x + teknium/OpenHermes-2.5 + sequelbox/Titanium4-DeepSeek-V4-Pro + VINAY-UMRETHE/Sonnet-Opus-4.5-4.6-Gemini-3.0-3.1-Pro-GPT-5-5.1-5.2-GLM-4.7-MiniMax-M2.1-DeepSeek-V3.2-High + ansulev/GPT-5.5-Thinking-Max-Distill-25k + Jackrong/Claude-opus-4.7-TraceInversion-5000x + Avtrkrb/combined-reasoning-opus-4.6-opus-4.7-kimi-k2.5-kimi-k2.6-glm-5.1 + sequelbox/Mitakihara2-DeepSeek-V4-Pro + open-r1/OpenR1-Math-220k + Roman1111111/gemini-3.1-pro-hard-high-reasoning + Jackrong/Kimi-K2.5-Reasoning-1M-Cleaned + lordx64/reasoning-distill-claude-opus-4-7-max

why-is-california

1 day ago

•

edited 1 day ago

[mention removed] Please finetune qwen3.5 9b to make it frakenstein [...]

one thing missing.. Le Chaton Fat distillation! /j

Tralalabs

1 day ago

@mlabonne @FlameF0X @armand0e @mradermacher @UnslothFan @yuxinlu1 @danielhanchen If someone can do it, please finetune a Frakenstein Qwen3.5 9B on all those datasets so it gets a 100% in every benchmark.

Tralalabs

1 day ago

@JustinLin610 @xiongwang @VoyagerXHF @bartowski @ollama Creators of Qwen, if you can do it, finetune it and upload it to Qwen org.

Tralalabs

1 day ago

@zjj1233 @harryhsing @xiangh @taesiri @lhjiang OMG pls

FlameF0X

1 day ago

•

edited about 24 hours ago

[mention removed] If someone can do it, please finetune a Frakenstein Qwen3.5 9B on all those datasets so it gets a 100% in every benchmark.

This is very off topic for this discussion, y'know?
Also, i got a i3-6006u, how am i supposed to even fine tune a 9B model. Sure, maybe Colab or Kaggle, but no. cuz i cant property use unsloth to fine tune a 9B model without hitting OOM.

Why not train the model yourself?

armand0e

about 7 hours ago

@Tralalabs stop mentioning everybody. It’s very annoying

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