xAI and MiniMax Models Now List Correctly in the Model Catalog

This is a fix, not a launch — xAI and MiniMax support already existed in the AI Gateway; it just wasn't visible.

What was broken

The gateway's model catalog endpoint (/v1/models) and its companion provider-availability endpoint (/v1/models/providers/list) each determine whether a given model shows as available with a per-provider check. Neither had a case for xAI or MiniMax at all — so even with a valid XAI_API_KEY or MINIMAX_API_KEY configured, those providers' models fell through to a default of "not available" and never appeared in the catalog. In practice, this meant the Gateway Portal's public model list only ever showed OpenAI, Workers AI, and Azure Foundry models, regardless of what was actually configured behind the scenes.

What's fixed

Both endpoints now include explicit handling for xAI and MiniMax, so a configured key for either provider makes its models show up correctly — matching how every other provider already worked.

A related but separate cluster of fixes, same week

Around the same time, a similar class of bug was found and fixed for OpenRouter specifically: OpenRouter models were also missing from both the public catalog filter and the provider list, for a different reason — OpenRouter is BYOK-only, so "available" there means a signed-in gateway customer has their own active OpenRouter key connected, not a platform-wide key. That required its own availability logic (checking the caller's identity and their stored key) rather than the simpler platform-key check used for xAI and MiniMax, and a companion fix so the Gateway Portal's playground actually forwards caller identity when building its model picker — without that, even a customer with a working OpenRouter key couldn't select an OpenRouter model in the playground UI. It's a related gap in the same area of code, but a distinct fix from the xAI/MiniMax one described above, not the same models or the same root cause.