> ## 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 status and code a call can return, what it means and what to do.

Errors have OpenAI's shape (Anthropic's on the Messages API): a status, a code and a message. Every answer carries an `x-os3-request-id` header: send it to [support](https://docs.os3compute.com/support.md) if you need help. Nothing is billed for a refused request.

| Status and code | What it means | What to do |
| - | - | - |
| `400` | The request is not valid for this model: a field DeepSeek's API refuses, a prompt that does not fit the context, or an image that cannot be read. | Read the error's message: it names what to change. |
| `400 unsupported_parameter`, `unsupported_tool`, `unsupported_content` | A field, tool or content type OS3 does not serve, such as `store: true` or web search. | Remove it. See [the model page](https://docs.os3compute.com/models/deepseek-v4-1-flash-1017.md). |
| `400 catalog_api_unsupported` | An API these models are not served on (Gemini, embeddings, `count_tokens`). | Use chat completions, Messages or Responses. |
| `401 invalid_api_key` | The key is missing, mistyped, revoked or expired. | Send your key as `Authorization: Bearer`, or create a new one in the portal. |
| `402 payment_required` | Your organisation has used up its credit and has no card on file. | Add a card or buy credit in **Usage**, then retry. |
| `402 card_required` | A training run, deployment or other GPU work was started without a card or enough credit. | Add a card or buy credit in **Usage**, then start it again. |
| `403 catalog_plan_required` | Your plan does not include the fast id (an earlier plan; every plan offered now does). | Use the normal id, or ask for a plan change in **Usage**. |
| `403 url_key_catalog` | The request used your setup URL's key, which cannot call OS3's models. | Send the key itself as `Authorization: Bearer`. |
| `403 catalog_region_refused` | OS3 limits your organisation to regions that do not include the United States, where these models run. | Ask [support](https://docs.os3compute.com/support.md) to add the United States. |
| `404 model_not_found` | The id is misspelt, or it is a withdrawn model. | Copy the id from [the model page](https://docs.os3compute.com/models/deepseek-v4-1-flash-1017.md). `GET /v1/models` lists the ones your key can call. |
| `413 request_too_large` | The request's body, with its images inline, is larger than OS3 accepts. | Send fewer or smaller images, or pass them by URL. |
| `429 rate_limit_exceeded` | Over your plan's requests or tokens a minute, for the key or the organisation. | Wait as long as the `retry-after-ms` header says. The OpenAI SDKs do this by themselves. |
| `429 tenant_busy` | Your organisation has its plan's number of requests in flight, or holds its half of the normal lane on a GPU server. | Retry after the `Retry-After` header's seconds (1 to 5), or run fewer at once. |
| `503 engine_busy` | OS3's GPUs are full. | Retry after the `Retry-After` header's seconds. |
| `503 model_unavailable` | OS3's GPUs for this model are down. No other model answers in its place. | Retry after 30 seconds. |

An error after a long 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.