> ## Documentation Index
> Fetch the complete documentation index at: https://docs.os3compute.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Errors

> Each code a call can return, and what to do.

Errors have OpenAI's shape (Anthropic's on the Messages API). Nothing refused is billed.

| Code | Meaning | Do |
| - | - | - |
| `400` | Not valid for this model, e.g. a prompt over the context. | Read the message. |
| `400 unsupported_*` | A field, tool or content OS3 does not serve. | Remove it. |
| `400 catalog_api_unsupported` | An API not served, e.g. embeddings. | Use chat completions, Messages or Responses. |
| `401 invalid_api_key` | Key missing, wrong, revoked or expired. | Send it as `Authorization: Bearer`. |
| `402 payment_required` | No credit and no card. | Add either in **Usage**. |
| `402 card_required` | Training started without a card or enough credit. | Add either, then start again. |
| `403 catalog_region_refused` | Your organisation excludes the United States. | Ask [support](https://docs.os3compute.com/support.md). |
| `404 model_not_found` | Unknown id. | `GET /v1/models` lists yours. |
| `413 request_too_large` | Body too large with its inline images. | Send fewer, smaller or URL images. |
| `429 rate_limit_exceeded` | Over your requests or tokens a minute. | Wait `retry-after-ms`. |
| `429 tenant_busy` | Too many requests in flight. | Wait `Retry-After` (1 to 5 seconds), or send fewer at once. |
| `503 engine_busy` | OS3's GPUs are full. | Wait `Retry-After`. |
| `503 model_unavailable` | The model is down. | Retry after 30 seconds. |

An error after an answer has started (status 200) comes as an `error` object in the body.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.