llm-openrouter library 0.6 adds multi-model support and streaming — developers can now switch between AI providers without rewriting code

Simon Willison released version 0.6 of llm-openrouter, a Python tool that lets developers access multiple AI language models (like Claude, GPT-4, Llama) through a single interface. The update adds the ability to stream responses in real-time, so outputs appear word-by-word instead of waiting for the entire answer.

Previously, switching between different AI providers meant rewriting your code for each one because they use different request formats. Now a developer can use the same three lines of code to call Claude one day and switch to an open-source model the next, reducing the friction of trying different AI services.

If you're a developer building AI applications, this means you can shop around for cheaper or faster models without lock-in — you could use an expensive model for quality-critical work and a cheaper one for routine tasks, all without refactoring. For AI hobbyists or small teams without dedicated DevOps, it removes a technical barrier to experimenting.