Instructions to use finnianx/LFM2.5-350M-THINKING with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use finnianx/LFM2.5-350M-THINKING with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="finnianx/LFM2.5-350M-THINKING",
	filename="LFM2.5-350M-THINKING.F16.gguf",
)

llm.create_chat_completion(
	messages = "No input example has been defined for this model task."
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use finnianx/LFM2.5-350M-THINKING with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf finnianx/LFM2.5-350M-THINKING:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf finnianx/LFM2.5-350M-THINKING:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf finnianx/LFM2.5-350M-THINKING:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf finnianx/LFM2.5-350M-THINKING:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf finnianx/LFM2.5-350M-THINKING:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf finnianx/LFM2.5-350M-THINKING:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf finnianx/LFM2.5-350M-THINKING:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf finnianx/LFM2.5-350M-THINKING:Q4_K_M

Use Docker

docker model run hf.co/finnianx/LFM2.5-350M-THINKING:Q4_K_M

LM Studio
Jan
Ollama
How to use finnianx/LFM2.5-350M-THINKING with Ollama:
```
ollama run hf.co/finnianx/LFM2.5-350M-THINKING:Q4_K_M
```

Unsloth Studio new

How to use finnianx/LFM2.5-350M-THINKING with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for finnianx/LFM2.5-350M-THINKING to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for finnianx/LFM2.5-350M-THINKING to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for finnianx/LFM2.5-350M-THINKING to start chatting

Docker Model Runner
How to use finnianx/LFM2.5-350M-THINKING with Docker Model Runner:
```
docker model run hf.co/finnianx/LFM2.5-350M-THINKING:Q4_K_M
```

Lemonade

How to use finnianx/LFM2.5-350M-THINKING with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull finnianx/LFM2.5-350M-THINKING:Q4_K_M

Run and chat with the model

lemonade run user.LFM2.5-350M-THINKING-Q4_K_M

List all available models

lemonade list

LFM2.5 350M Base

Finetuned on the following datasets:

TeichAI/lordx64-claude-opus-4.7-max-cleaned
Jackrong/DeepSeek-V4-Distill-8000x

Training Settings

Trained with the ChatML chat format
Effective batch size: 16
Training epochs: 2

LoRA Settings

16-bit LoRA
r = 16
lora_alpha = 32

Warning

This model sometimes gets stuck in thinking loops, mostly on very simple prompts.

Sample Output

Prompt:
Is 9.11 bigger than 9.9?

We need to determine if 9.11 is bigger than 9.9. The question asks "Is 9.11 bigger than 9.9?" This is a simple comparison. We can check if 9.11 is greater than 9.9 by comparing the values. Since 9.11 is less than 9.9, the answer is no. So the answer is "No".

Answer: No.

Explanation:

9.11 is less than 9.9. Therefore, 9.11 is not greater than 9.9.

LFM2.5-350M-THINKING : GGUF

This model was finetuned and converted to GGUF format using Unsloth.

Example usage:

For text only LLMs: llama-cli -hf finnianx/LFM2.5-350M-THINKING --jinja
For multimodal models: llama-mtmd-cli -hf finnianx/LFM2.5-350M-THINKING --jinja

Available Model files:

LFM2.5-350M-THINKING.Q8_0.gguf
LFM2.5-350M-THINKING.F16.gguf
LFM2.5-350M-THINKING.Q4_K_M.gguf This was trained 2x faster with Unsloth

Downloads last month: 308

Safetensors

Model size

0.4B params

Tensor type

BF16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for finnianx/LFM2.5-350M-THINKING

Base model

LiquidAI/LFM2.5-350M-Base

Quantized

(1)

this model

Quantizations

1 model

finnianx
/

LFM2.5-350M-THINKING