Errors

Status	Meaning
`400`	Bad request — unknown model, invalid body, or oversized non-stream `max_tokens`
`402`	Payment required — see the `PAYMENT-REQUIRED` header for price, `payTo`, and memo
`413`	Request body too large — limit is 2 MB
`429`	Too many requests — queue limit reached
`503`	Service busy — retry with backoff

All errors use the OpenAI error format:

{
  "error": {
    "message": "...",
    "type": "...",
    "code": "..."
  }
}

Empty answers are not an error#

A 200 response with empty content is not a failure state on BridgeNode's side — it means the model consumed its entire max_tokens budget on reasoning and had nothing left for the visible answer. This is common with thinking models when max_tokens is set too low.

Empty answers are not refunded (the provider returned 200; the service was provided).
Fix: increase max_tokens to at least 200 and submit a new paid request.

See Models & Pricing → Reasoning models for the full explanation.

Refunds#

If the provider fails before any content is delivered, the payment is refunded automatically via a reverse USDC transfer. See Payment Flow for the settlement mechanics.

Updated Aug 14, 2026