🤖 HarDojo
Log in Sign up

Local Models with Ollama

Run open models (Llama, Qwen, Hermes) on your own machine.

🧪 Try it in the Playground 🎯 Practice this path

Ollama runs open-source models locally — no cloud, no API key, no data leaving your machine.

Install & run

# macOS / Linux / Windows (WSL2)
curl -fsSL https://ollama.com/install.sh | sh

ollama run llama3.1      # chat with Llama
ollama pull qwen2.5-coder # a strong open coding model

Ollama exposes a local API (http://localhost:11434) that any tool can use.

Use it with your agents

# Aider with a local model
aider --model ollama/llama3.1

# Codex CLI can also point at compatible endpoints
codex --config '{"model_provider":"ollama","model":"qwen2.5-coder"}'

Choosing a model

| Model | Strengths | | --- | --- | | llama3.1 / llama3.3 | Solid generalist, fast | | qwen2.5-coder | Strong at code for its size | | deepseek-coder | Excellent code, chatty | | hermes3 | Built for tool use / function calling | | codestral | Mistral's coding specialist |

Trade-offs

  • Pro: privacy, free, offline, no rate limits.
  • Con: smaller context, slower, and weaker on the hardest multi-file tasks than frontier cloud models.
A common pro workflow: use a local model for autocomplete and simple refactors, and a cloud model for big agent tasks. Best of both.
";