MODEL_UNAVAILABLE

What triggered it

VibeCord routes generations through OpenRouter, which fronts dozens of underlying model providers (Anthropic, MiniMax, Z.ai, Moonshot, etc.). MODEL_UNAVAILABLE (HTTP 503) means OpenRouter responded but said “the upstream provider is currently unreachable.”

Common causes:

• The provider is doing maintenance (usually 15-30 minutes).
• Rate-limit at the upstream — provider hit their own quota with OpenRouter.
• Region-specific outage — e.g. one model’s hosting region went down.
• Model was deprecated — rare, but providers sometimes pull old model versions without notice.

Your in-progress work is saved

If a generation was already running and the model became unavailable mid-stream, we keep everything emitted so far. Your project is intact and you can iterate from where it stopped.

What to try

1. Switch models. Settings → Model. If MiniMax M2.5 is down, try Z.ai GLM-5 (quality tier) or Kimi K2.5 (reliability tier).

2. Wait 5-10 minutes and retry. Most upstream blips clear quickly.

3. Check the status page. We surface OpenRouter incidents within 60 seconds of detecting them.

4. Check OpenRouter status. status.openrouter.ai shows per-provider health. If your model is red there, switch.

Back to the app

How model selection works

Every plan starts on a sensible default model. You can override it in Settings → Model; the override persists per project.

When to contact support

If MODEL_UNAVAILABLE keeps firing across multiple models for more than 30 minutes:

• Drop a note in Discord
• Or email support@vibecord.dev

This usually means OpenRouter itself is degraded and we’re already on it — but reports help us prioritize.