Exploring Open-Source AI: Running Local LLMs on Ubuntu

To Nha Notes | Feb. 20, 2025, 9:40 p.m.

🔍 1. Discover & Browse Local LLMs

These platforms allow you to explore and download open-source AI models:

  • Hugging Face Model Hub – The biggest repository for LLMs like LLaMA, Mistral, Falcon, and more.
  • TheBloke on Hugging Face – Optimized quantized models for running efficiently on local hardware.
  • ModelScope – A model hub focused on Chinese and multilingual models.
  • CivitAI – AI models and LLMs focused on image generation but expanding into text models.

📥 2. Download & Run Local LLMs

These tools allow you to run local models efficiently on Ubuntu:

Easy-to-Use LLM Runtimes

  1. Ollama – Simplest way to run local LLMs with ollama run mistral.
  2. LM Studio – GUI application for discovering and running local models.
  3. GPT4All – Cross-platform LLM runner with chat UI.

Command-Line & Web UIs

  1. Text Generation Web UI – Web UI for running models like LLaMA, Mistral, and GPT-J.
  2. KoboldCpp – Optimized for running large models using CPU/GPU.
  3. LLM CLI – Simple terminal-based local LLM runner.

For Advanced GPU Users

  1. LMDeploy – NVIDIA’s optimized LLM deployment toolkit.
  2. vLLM – High-performance LLM inference for GPUs.
  3. ExLlama – Optimized for running quantized LLaMA models.

🏃 3. Running a Local LLM on Ubuntu

Example: Using Ollama to run Mistral locally

# Install Ollama curl -fsSL https://ollama.com/install.sh | sh # Download and run Mistral-7B ollama run mistral

Example: Using LM Studio

  1. Download LM Studio.
  2. Open LM Studio and browse models.
  3. Download and run any model with a GUI.

Would you like recommendations based on your hardware specs (CPU/GPU)? 😊