LLM API Error Codes
The build is broken and you have pasted a wire error into a search box. This is the exact body each provider sends, why it happens there and not on its neighbor, and a fix you can run — not a retelling of the provider's own docs.
What does this LLM API error mean and how do I fix it?
5 provider-specific error pages are registered here, each with the exact wire body, why it happens on that provider and not its neighbors, and a fix as executable code — not paraphrased provider docs. Errors whose fix is identical across every provider (bad JSON, a missing key, an oversized payload) are answered inline below instead of getting a page each.
Provider-specific errors
| Provider | Error | HTTP | Retryable | Status |
|---|---|---|---|---|
| Anthropic | max_tokens: Field required | 400 | No | Preview (draft) → |
| OpenAI | Parameter locked to a fixed value | 400 | No | Preview (draft) → |
| Anthropic | overloaded_error | 529 | Yes | Preview (draft) → |
| OpenAI | rate_limit_exceeded | 429 | Yes | Preview (draft) → |
| Anthropic | rate_limit_exceeded | 429 | Yes | Preview (draft) → |
Same fix on every provider — no page needed
These errors have the same answer regardless of which provider sent them, so publishing one page per provider would just be the same sentence eleven times.
| Error | Fix |
|---|---|
bad_json | The request body is not valid JSON, or is missing a required top-level field — validate it against the provider's schema before sending, on every provider. |
auth_missing | No API key was sent at all — check the header name (Authorization: Bearer vs. x-api-key differ by provider) and that the environment variable is actually loaded. |
payload_too_large | The request body exceeds the provider's size ceiling — split the payload, or drop attached content that isn't needed for this call. |
Full error taxonomy: /errors/data.json — cite this dataset directly.