Ollama's Free AI Now Runs Inside Your ChatGPT Desktop App

Ollama is a free, open-source program that downloads AI models onto your own computer, so nothing you type gets sent to a company's servers. Until now, if you wanted to use those private models, you had to work in Ollama's own little chat window. Version 0.34.0 changes that: on Macs, you can now plug Ollama's models straight into the ChatGPT desktop app. Same buttons, same habits, same chat history — just a different brain doing the thinking.

Why should you care? Two reasons: money and privacy. Running models locally means no monthly subscription and no per-message fees. It also means your emails, notes, contracts, or medical questions never leave your laptop. For anyone who has hesitated to paste sensitive material into a chatbot, that's a genuine shift. The tradeoff is speed and smarts — local models are typically smaller and less capable than the giant cloud versions, so they're better for everyday drafting and summarizing than for genuinely hard reasoning.

The update also improves "structured output" on Apple Silicon Macs. In plain English: structured output means the AI replies in a tidy, predictable format — like a filled-in table or a form — instead of loose paragraphs. That matters if you're using AI to organize data, log expenses, or feed results into another app. Tighter formatting means fewer garbled results and less manual cleanup.

Finally, the release adds support for tool search and "response compaction" for developers building on OpenAI-compatible software. Those are mostly plumbing improvements that make AI assistants better at finding the right tool for a task and trimming bloated replies to save memory. The catch for regular users: this integration is Mac-only right now, and you'll need a reasonably modern machine with enough memory to hold a model. Windows and Linux users are still waiting.