Any model you like.
Local through Ollama. Multimodal when you want eyes on the page. Switch from the picker in chat.
Point dassi at localhost:11434 — nothing leaves your machine.
Local through Ollama. Multimodal when you want eyes on the page. Switch from the picker in chat.
Connect your local model on any plan. Compare plans →
Repeated context is reused, known sites aren't re-read. Same job, fewer tokens — on your hardware, that's speed. How it works →
ollama pull qwen3:8b OLLAMA_ORIGINS=chrome-extension://bjcngahpcjeililljmfegmlanlpgibdi ollama serve Then in dassi: ⋮ → Settings → Models → API keys → Custom, base URL http://localhost:11434, Save. Full guide with every screenshot →