Build with your API key
Build a website with Ollama (free local AI)
Facts last verified: . Provider pricing and free tiers change — check their site for current numbers.
Yes — and it costs nothing: Ollama runs open models like Llama and Qwen 2.5 Coder locally on your own hardware, so there is no API key, no per-token bill, and no data leaving your machine. Install Ollama, pull a model, point OpenThorn at your local endpoint, and build.
Get your key
Official pricing
Free tier
Entirely free — models run on your own hardware. The only requirement is enough RAM/VRAM for the model you choose.
Recommended models
- Qwen 2.5 Coder — The best local coding model for its size — recommended default.
- Llama 3.2 — Light and fast on modest hardware.
- Mistral — Solid all-rounder that runs on most machines.
How to build your website
- Install OllamaDownload Ollama from ollama.com for Windows, macOS, or Linux and install it — it runs a local server on port 11434.
- Pull a coding modelRun `ollama pull qwen2.5-coder` in a terminal. A 7B model needs roughly 8 GB of RAM; larger variants produce better code.
- Connect Ollama to OpenThornSign up free at openthorn.app, open the Providers page, choose Ollama, and point it at your local endpoint. No API key required.
- Describe your website and ship itCreate a project, pick your local model, and describe the site. Everything runs on your machine; deploy to Netlify or export when ready.
What it costs
$0 — inference runs on your own hardware. Larger models produce better websites but need more RAM/VRAM (a 7B coder model runs on most modern laptops).
Common questions
Can I really build a website with AI completely free and offline?
Yes. Ollama runs open models locally — no API key, no token bills, no cloud. OpenThorn's platform is free, so the entire build costs $0. Quality depends on the model size your hardware can run; local models are weaker than cloud flagships but capable for simpler sites.
What hardware do I need for Ollama?
A 7B coding model like Qwen 2.5 Coder runs on most modern laptops with 8 GB+ RAM. More RAM or a GPU lets you run larger variants, which noticeably improves generated code quality.
Is local AI private?
Yes — with Ollama your prompts and generated code never leave your machine. This is the strongest privacy option among all providers OpenThorn supports.
New here? Read how to build a website with AI using your own key, how to get an AI API key, or compare model pricing. Other providers: OpenAI · Anthropic · Google Gemini · DeepSeek · Mistral AI · Groq · Together AI · xAI · Cohere · Perplexity · OpenRouter · Fireworks AI · Cerebras · Azure OpenAI · Amazon Bedrock · NVIDIA NIM