Errors

Errors return a JSON body with a code and a message. The message says what happened and what to do.

Error body

{
  "error": {
    "message": "No node is serving deepseek/deepseek-v4.1-flash right now; retry shortly",
    "type": "rate_limit_error",
    "code": "model_not_available"
  }
}

HTTP	Code and message	What to do
400	invalid_request "Invalid request: `messages` is required"	Fix the request. The message names the field.
401	invalid_api_key "API key not recognised"	Check the key was copied in full, or create a new one.
401	sail_signature_invalid "This Sail key works only from the machine that signed in with it; run sail login on this machine"	Run sail login on the machine that sends the request. If the message says the signed request was already used, the same request was sent twice: send a new one.
401	sail_clock_skew "The request's time is more than 5 minutes from the gateway's; check this machine's clock"	Set the machine's clock to the network time, then retry.
402	insufficient_credits "Workspace has $0.00 in credits; add credits on the billing page"	Add credits on the billing page.
402	key_limit_reached "This key has reached its monthly limit of $5.00; raise the limit on the keys page"	Raise the key's monthly limit, or use another key.
403	region_not_allowed "This key is not allowed in region {region}; allowed: {regions}"	Send the request to an allowed region, or change the key's rules.
403	workspace_suspended "This workspace is suspended; contact hello@safeailabs.ai"	Write to us. Every key in the workspace is refused until the suspension is lifted.
403	sail_key_requires_sail "This key works only from Sail; create an API key at https://safeailabs.ai/keys for other clients"	Create an API key on the keys page for other clients. A key made by sail login works only from Sail.
404	model_not_found "The model `deepseek-r2` does not exist or is not available"	Check the id against `/v1/models`. A model not served in the region you called, or not allowed by the key's rules, answers the same way.
404	not_found "No route for POST /v1/embeddings"	See Compatibility for what is served.
413	context_length_exceeded "This request needs about 40000 tokens of context, over the model's limit of 32768"	Shorten the conversation, or pick a model with a longer context.
429	rate_limited "Rate limit of 5000 requests / min reached; retry after 12 s"	Wait for `Retry-After`, then retry.
429	rate_limited "This workspace's API plan allows 8 requests at once; none finished within 60 s"	The workspace's API plan had every place in use for 60 s. Wait for `Retry-After`, then retry. See Plans.
429	too_many_attempts "Too many wrong API keys from this address; retry in a minute"	Check the key, then retry after a minute.
429	model_not_available "No node is serving deepseek/deepseek-v4.1-flash right now; retry shortly"	Retry after `Retry-After`, or try another model. The request was not sent anywhere else.
503	region_unavailable "This region is not serving requests right now; retry later or use another region"	Retry after `Retry-After`, or call another region's host. The request was not sent to any other region.
503	service_unavailable "Keys cannot be checked right now; retry shortly"	Retry after `Retry-After`.

429 and 503 carry a Retry-After header in seconds. When every node serving the model is busy the answer is 429 with a Retry-After of 1 second and the number 429 as its code. A request is never retried in another region on your behalf. If you want a fallback region, call it yourself.

NextRate limits Retention

These docs as plain text for coding agents: llms.txt