Home → Help

Model not found — when the model name is almost right

What you are seeing

• The model `...` does not exist or you do not have access to it
• 404 model_not_found
• {"error":{"code":"model_not_found"}}

Why it happens

Naming is the usual culprit, and it is genuinely hard to guess. Within a single family you will find versions written with a dot in one release and a hyphen in the next — a name that reads identically to a human is a different string to the API.

The second cause is access rather than naming. On gateways where keys belong to a plan or group, a model can exist on the platform yet sit outside what your key is allowed to call. The message often does not distinguish the two cases, which sends people looking for a typo that is not there.

Confirm it is this

Ask the API what your key can see, rather than trusting a documentation page:

curl -s 'YOUR_BASE_URL/models' \
  -H 'Authorization: Bearer YOUR_KEY' \
  | grep -o '"id":"[^"]*"'

If the model is missing from that list, no amount of fixing the spelling will help — the key cannot reach it.

How to fix it

1. Copy the ID, never retype itTake the string straight from the models listing. Dots against hyphens, and version suffixes with dates, are the two things people get wrong most often.
2. Check your key's scopeIf the model appears in public documentation but not in your own listing, the problem is what the key is allowed to reach. A key for one plan usually cannot call another plan's models.
3. Watch for aliases that disappearConvenience aliases that point at whatever is current can be retired without notice. Pinning the fully qualified ID is more verbose and considerably more stable.

Related

Last checked 2026-08-24. Written from problems diagnosed on a live OpenAI-compatible gateway, not collected from other sites.