Error codes - Together AI docs

Code	Cause	Solution
400 - Invalid Request	The request is misconfigured, the input token count plus `max_tokens` exceeds the model’s context length, or a request to the dedicated model inference management API includes an unknown field.	Check that the request is valid JSON and your API key is correct. Use the prompt format the model expects, which differs between Mistral and Llama models. If the request exceeds the context length, lower `max_tokens`. For a chat model, you can also set `max_tokens` to `null` to let the model decide when to stop generating. If a management API request returns an `unknown field "..."` error, remove the named field.
401 - Authentication Error	The API key is missing or invalid.	Check that you’re using the correct API key and passing it correctly.
402 - Payment Required	Your organization’s balance has reached its spend cut-off ($0 for prepaid accounts), or the model requires a positive credit balance.	Add credits to resume service. If you’ve already paid, wait up to 5 minutes for your balance to update, then retry.
403 - Forbidden	Your account can’t access the requested model or feature. For example, your plan doesn’t include the model, the model requires a credit card on file, your organization has turned off passthrough models, or the account is blocked.	The error message names the reason. Upgrade your plan or add a payment method in your billing settings, or ask an organization admin to turn on passthrough models. If your account is blocked, contact support.
404 - Not Found	The endpoint URL or model name is invalid.	Check that the endpoint URL matches the API reference and that the model you’re querying is available.
413 - Request Entity Too Large	The request body is larger than the endpoint accepts. Limits vary by endpoint.	Reduce the request size. Pass large images or audio as a URL instead of embedding them in the request body. Direct audio uploads are capped at 80 MB (see transcription limits).
429 - Too Many Requests	Together is limiting serverless requests during high demand, or a deployment on a dedicated endpoint would exceed your project’s or organization’s GPU quota for its GPU type.	For serverless rate limiting, reduce your request rate, spread out bursts, and retry with exponential backoff. For GPU quota, lower the requested replica count, stop unused deployments on that GPU type, or contact support to raise the quota. If the platform itself is out of capacity for that GPU type, retry later.
500 - Server Error	An unknown server error occurred.	Wait briefly, then retry. If the error persists, contact support.
503 - Service Unavailable	The service is temporarily unavailable, for example during high demand.	Wait briefly, then retry with exponential backoff. If the error persists, contact support.
504 - Timeout	The request didn’t complete in time.	Wait briefly, then retry. If the error persists, contact support.
524 - Cloudflare Timeout	The connection timed out at the network edge.	Wait briefly, then retry. If the error persists, contact support.
529 - Overloaded	The model has reached its maximum replica count and can’t take more requests.	Wait briefly, then retry with exponential backoff. If the error persists, contact support to raise the model’s maximum replica count.

If you see an error code that isn’t listed here, or a solution doesn’t resolve the error, contact support.

⌘I