BerriAI/litellm · error · RateLimitError

{custom_llm_provider.capitalize()}Exception: Rate Limit Errr

Error message

{custom_llm_provider.capitalize()}Exception: Rate Limit Errror - {error_str}

What it means

LiteLLM raises RateLimitError when the provider's error string contains the literal 'token_quota_reached', i.e. your account or key exhausted its token quota for the period. Note the message contains a typo ('Rate Limit Errror' with three r's), so string-matching on the message is unreliable — catch the exception type instead.

Source

Thrown at litellm/litellm_core_utils/exception_mapping_utils.py:733

            error_str += "XXXXXXX" + '"'

        raise AuthenticationError(
            message=f"{custom_llm_provider.capitalize()}Exception: Authentication Error - {error_str}",
            llm_provider=custom_llm_provider,
            model=model,
            response=getattr(original_exception, "response", None),
            litellm_debug_info=extra_information,
        )
    elif ExceptionCheckers.is_error_str_context_window_exceeded(error_str):
        raise ContextWindowExceededError(
            message=f"{custom_llm_provider.capitalize()}Exception: Context Window Error - {error_str}",
            model=model,
            llm_provider=custom_llm_provider,
            response=getattr(original_exception, "response", None),
            litellm_debug_info=extra_information,
        )
    elif "token_quota_reached" in error_str:
        raise RateLimitError(
            message=f"{custom_llm_provider.capitalize()}Exception: Rate Limit Errror - {error_str}",
            llm_provider=custom_llm_provider,
            model=model,
            response=getattr(original_exception, "response", None),
        )
    elif "The server received an invalid response from an upstream server." in error_str:
        raise litellm.InternalServerError(
            message=f"{custom_llm_provider.capitalize()}Exception - {original_exception.message}",
            llm_provider=custom_llm_provider,
            model=model,
        )
    elif "model_no_support_for_function" in error_str:
        raise BadRequestError(
            message=f"{custom_llm_provider.capitalize()}Exception - Use 'watsonx_text' route instead. IBM WatsonX does not support `/text/chat` endpoint. - {error_str}",
            llm_provider=custom_llm_provider,
            model=model,
        )
    elif hasattr(original_exception, "status_code"):

View on GitHub (pinned to 6c2dcb801b)

Solutions

Wait for the quota window to reset, or request a quota increase / top up the key.
Switch to a different API key or provider via litellm.Router fallbacks.
Track token usage (litellm's success callbacks) and alert before hitting the cap.
If you control the gateway, raise the token quota for that key.

Example fix

# before
try:
    resp = litellm.completion(model=..., messages=msgs)
except litellm.RateLimitError as e:
    if "Errror" in str(e):  # fragile string matching
        ...

# after
try:
    resp = litellm.completion(model=..., messages=msgs)
except litellm.RateLimitError:
    # handle by type; message text ('Rate Limit Errror') may change between versions
    time.sleep(60)
    resp = litellm.completion(model=..., messages=msgs)

Defensive patterns

Strategy: retry

Type guard

import litellm

def is_token_quota_error(e: BaseException) -> bool:
    return isinstance(e, litellm.RateLimitError) and "token_quota_reached" in str(e)

Try / catch

try:
    resp = litellm.completion(model=m, messages=msgs)
except litellm.RateLimitError as e:
    if "token_quota_reached" in str(e):
        wait_until_quota_reset()  # or rotate to another key/model
    else:
        raise

Prevention

Monitor cumulative token usage with litellm success callbacks and alert near the quota.
Do not match on the message text (it contains a typo and may change); match the exception type plus the provider token.
Keep a secondary API key/model configured for quota exhaustion.

When it happens

Trigger: A completion() call against a provider/gateway (e.g. Cloudflare AI Gateway or an internal gateway) that returns 'token_quota_reached' in its error body once the configured token budget is spent.

Common situations: Teams with per-key token budgets on a gateway, free-tier keys with daily caps, CI loops burning through quota, or shared proxy keys where one workload exhausts the pool.

Related errors

AI-assisted analysis of BerriAI/litellm@6c2dcb801b (2026-08-15). Data as JSON: /api/errors/55e431f48d47a6e5. Report an issue: GitHub.