Status

The moving parts behind verified output, and where to check each one is alive.

Is my call failing — me or the service?

A quick triage:

• 401 — bad or missing key. Verified output uses your normal maxmodel.com gateway token (Authorization: Bearer <key>).
• 400 sources_too_large / too_many_cases — you’re over a documented limit, not an outage.
• 5xx / timeouts — most often the underlying model the gateway routed to, not the verification layer. Retry, or try a faster model (see Models); a responses-endpoint model like gpt-5.5-pro can legitimately take 10–20s before it’s “failing.”
• See Errors for the full HTTP + typed-error reference.

The verification step itself is deterministic local code — when something is slow or flaky, it’s almost always the model call in front of it, not the check.