Prompt Router Puts a Stable Name on Unstable Models

Field notes

The service preserves the OpenAI chat-completions shape while changing the selection vocabulary: developers send basic, pro, or max rather than a provider-specific snapshot identifier. The published tier table gives concrete operating boundaries, including 32,768 to 200,000-token context windows and maximum outputs from 4,096 to 32,768 tokens. A linked changelog illustrates the problem it targets with deprecations, renames, and preview replacements. The page also states that streaming and tools remain compatible, though those capabilities are not independently demonstrated in the extract.