Model Availability & Reliability

Overview

Genesis Writer routes to 20+ models from every major provider. Those providers occasionally have outages, hit capacity, or retire a model with little notice. You don't have to track any of that — the platform detects trouble and keeps you writing, automatically. This page explains what you'll see when something goes wrong upstream.

Models Under Maintenance

If a model starts failing repeatedly — or a provider pulls it entirely — Genesis Writer takes it offline on its own. In the model selector, an unavailable model shows up greyed out with a small wrench icon and a tooltip like “This model is temporarily unavailable — we're looking into it.” You can't pick it until it recovers, which it does automatically once the provider is healthy again.

Automatic Fallback

If your chosen model fails during a generation — a timeout, a rate limit, a transient server error — Genesis Writer quietly retries the request on a healthy backup model so it still completes. This happens server-side, so most of the time you won't even notice a hiccup occurred; you just get your prose.

When a Model Can't Respond

If a model simply can't answer — it's fully down, or it declined the request — you'll get a calm, plain-language message instead of a cryptic error, usually suggesting you pick a different model from the menu. You're only ever charged for the tokens actually used, so a failed or stopped generation won't drain your balance.

The Default Model Is Health-Aware

New sessions start on a sensible default model (currently GPT-5.6 Luna on the Free plan). That default is checked for health before it's handed to you — if the usual default is having problems, Genesis Writer automatically serves the next healthy Free-plan model instead, so you're never dropped onto a broken default.

What to Do If a Model Is Unavailable

If the model you want is greyed out or a generation fails, just pick another model from the selector in the toolbar — switching takes one click and doesn't touch any text you've already generated. Outages almost never hit every model at once, so there's always a capable alternative a click away. If you were mid-scene, switch models and use Continue to pick the thread back up.