So you want to use OpenRouter?
So you want to use OpenRouter? One of OpenRouter's selling points is that it "handles fallbacks automatically and picks the most cost-effective option for each request", so you can call a single API endpoint for a model and get routed to the best available backend provider. Mohamed Moustafa points out a whole set of ways that this can cause you problems. Different providers run different serving…
If you're considering using OpenRouter for your AI model needs, there are a few important factors to keep in mind. One of OpenRouter's main selling points is its ability to automatically handle fallbacks and select the most cost-effective option for each request. This means you can simply call a single API endpoint for a model, and OpenRouter will route your request to the best available backend provider.
However, Mohamed Moustafa warns that this feature can sometimes lead to problems. Different providers utilize varying serving software with distinct optimizations and settings. As a result, the same OpenRouter endpoint can deliver model requests that behave differently depending on the provider. Furthermore, certain providers may lack vision capability for vision models, and the processing of the reasoning effort option can vary as well.
To mitigate these issues, OpenRouter offers the provider.only option, which allows you to control which provider is routed to. If you want to see the list of available providers for a specific model ID, you can use the /endpoints method.
Simon Willison, in a link post dated 11th September 2026, highlights these considerations when using OpenRouter. If you'd like to receive regular updates on the most important developments in Large Language Models (LLMs), you can sponsor the source for $10/month and receive a curated email digest.
Written by urgent.news from Simon Willison's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
