We've made reliability improvements for Claude models in our LeMUR framework.

We've made adjustments to our infrastructure so that users should see fewer timeout errors when using our Nano tier with some languages.