DeepInfra inference errors: separate rate limit, billing, and model errors

TL;DR

400 with contextlengthexceeded means the prompt exceeded the model's Context: A bare HTTP status is not enough to decide retry behavior.

The error

separate rate limit, billing, and model errors

Fix

1. 400 with contextlengthexceeded means the prompt exceeded the model's Context: A bare HTTP status is not enough to decide retry behavior. Expected: you can confirm this matches your situation.

2. Re-run the original operation and confirm the error is gone. Expected: no error, normal output.

When to use

• You hit this exact error with DeepInfra inference errors.
• The symptom matches: separate rate limit, billing, and model errors.

When NOT to use

• A different error message from DeepInfra inference errors; the cause here is specific to this error.
• Unrelated DeepInfra inference errors issues (different feature, different failure).
• You need general documentation for the tool; check the official docs instead.

Compatibility

Reported against DeepInfra inference errors. Source: https://github.com/revisionlabsai/revidotnet/blob/HEAD/ReviDotNet.Core/Docs/inference-errors.md.

Variant phrasings

separate rate limit, billing, and model errors

Why it happens

trim, do not retry. 400 modelnotfound and 404 both mean the model ID is wrong; check the model list. 429 without a billing code is a real rate limit: back off. 429 with insufficientquota or creditbalance_exhausted is a billing problem: retrying burns time, top up instead. 408, 409, 498, and 5xx are transient: retry with backoff. 401 is authentication: fix the key.

Edge cases

• If your error message differs even slightly, this is probably a different issue; search the exact text.
• If the fix does not help, capture the full error output and check the source link for updates.

Source

https://github.com/revisionlabsai/revidotnet/blob/HEAD/ReviDotNet.Core/Docs/inference-errors.md