Qwen3.6-35B-A3B on my laptop drew me a better pelican than Claude Opus 4.7

1. 1
   Identify Your Local AI Task

   Determine a specific AI task (e.g., text generation, image generation, summarization) where you currently use a cloud model or could benefit from local processing. Think about tasks where privacy, latency, or cost are critical.
2. 2
   Research Suitable Local Models

   Search for open-source or locally deployable models that excel at your identified task. Look for models optimized for edge devices or specific hardware, such as variations of Qwen, Llama, or Stable Diffusion that can run on consumer hardware.
3. 3
   Choose a Local Inference Platform

   Select a platform or framework for running local models. Popular choices include Ollama (for LLMs), LM Studio, or direct inference engines like `llama.cpp` or `diffusers` depending on the model type.
4. 4
   Install Platform and Download Model

   Set up your chosen local runner (e.g., install Ollama). Then, download a target model (e.g., a Qwen variant) to your local machine using the platform's commands or direct download.
5. 5
   Run and Evaluate Performance

   Execute your chosen AI task with the local model. Compare its output quality, generation speed, and resource usage against cloud-based alternatives (if you have them) or against your expectations for the task.
6. 6
   Leverage Local AI Benefits

   Based on your evaluation, integrate the local model into your workflow or applications. Capitalize on benefits like enhanced data privacy, reduced latency, and lower operational costs compared to continuous cloud API calls.

# Install Ollama (if not already installed)
# curl -fsSL https://ollama.com/install.sh | sh

# Pull a Qwen model (e.g., Qwen:7B or Qwen:14B)
ollama pull qwen:7b

# Run the model and ask it for a creative description
ollama run qwen:7b "Describe a pelican in flight over a stormy sea, in a poetic style, focusing on its wingspan and grace."

• Articlesimonwillison.net