Llama

Use Llama for one-click model downloads, a built-in chat, and a local OpenAI-compatible API that other apps can use, with no command-line setup.

Llama runs GGUF models with llama.cpp and keeps them in the standard Hugging Face cache.

Installation

Download and install Llama from llama.app:

macOS (Apple Silicon, macOS 15 or later)
Windows 11 (x64 and ARM64)

Downloading Models

To install LFM3-1.7B, open this link in your browser:

llama://install?repo=LiquidAI/LFM3-1.7B-GGUF

Llama picks a quantization that fits your computer’s memory. The link works for any LFM GGUF repository: replace the repository name. See the Models page for all available GGUF models.

Using the Chat Interface

Open the model in Llama
Click Chat with model to open the chat in your browser
Start chatting

The model loads on the first message.

Using the API

Llama serves every installed model through a local OpenAI-compatible server at http://localhost:9931/v1 (the port is configurable in settings). Models load on the first request. Get the model IDs before sending requests:

curl http://localhost:9931/v1/models

IDs take the form org/repo:QUANT, for example LiquidAI/LFM3-1.7B-GGUF:Q4_K_M. Use the OpenAI Python client:

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:9931/v1",
    api_key="not-needed"
)

response = client.chat.completions.create(
    model="LiquidAI/LFM3-1.7B-GGUF:Q4_K_M",
    messages=[
        {"role": "user", "content": "What is machine learning?"}
    ],
    temperature=0.1,
    max_tokens=2048,
    extra_body={"top_k": 50, "repeat_penalty": 1.1},
)
print(response.choices[0].message.content)

Streaming Responses

stream = client.chat.completions.create(
    model="LiquidAI/LFM3-1.7B-GGUF:Q4_K_M",
    messages=[
        {"role": "user", "content": "Tell me a story."}
    ],
    stream=True
)

for chunk in stream:
    if chunk.choices[0].delta.content is not None:
        print(chunk.choices[0].delta.content, end="")

You can also use curl to interact with the server:

Curl request example

curl http://localhost:9931/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "LiquidAI/LFM3-1.7B-GGUF:Q4_K_M",
    "messages": [{"role": "user", "content": "Hello!"}],
    "temperature": 0.1,
    "top_k": 50,
    "repeat_penalty": 1.1
  }'

Using with Agents

LFM3-1.7B is post-trained for agent harnesses such as Pi, Hermes Agent, and OpenClaw. To run it, follow Run local agents with LFMs and point the harness at http://localhost:9931/v1. For harder tasks, use LFM2.5-24B-A2B from Llama’s catalog.

Models

Fine-tuning

Edge Inference

GPU Inference

Cloud Inference

Guides

Help

Installation

Downloading Models

Using the Chat Interface

Using the API

Streaming Responses

Using with Agents

​Installation

​Downloading Models

​Using the Chat Interface

​Using the API

​Streaming Responses

​Using with Agents

Installation

Downloading Models

Using the Chat Interface

Using the API

Streaming Responses

Using with Agents