diff --git a/litellm/proxy/pass_through_endpoints/streaming_handler.py b/litellm/proxy/pass_through_endpoints/streaming_handler.py
index 19d8b063dd7..6cda69bd8fe 100644
--- a/litellm/proxy/pass_through_endpoints/streaming_handler.py
+++ b/litellm/proxy/pass_through_endpoints/streaming_handler.py
@@ -12,6 +12,7 @@ from litellm.litellm_core_utils.asyncify import asyncify
 from litellm.litellm_core_utils.core_helpers import bind_budget_reservation_to_callbacks
 from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
 from litellm.litellm_core_utils.logging_worker import GLOBAL_LOGGING_WORKER
+from litellm.llms.anthropic.common_utils import AnthropicError
 from litellm.proxy._types import PassThroughEndpointLoggingResultValues
 from litellm.proxy.common_request_processing import ProxyBaseLLMRequestProcessing
 from litellm.proxy.common_utils.sse_keepalive import split_complete_sse_frames
@@ -297,6 +298,7 @@ class PassThroughStreamingHandler:
         from litellm.llms.anthropic.pass_through.messages.streaming_iterator import (
             _is_message_stop_chunk,  # pyright: ignore[reportPrivateUsage]  # both native stream paths share terminal-event detection
             _is_provider_error_chunk,  # pyright: ignore[reportPrivateUsage]  # provider errors must not become cache evidence
+            parse_anthropic_error_event,
         )
 
         # Transport reads can split event names and JSON payloads. Recognize terminal
@@ -312,6 +314,24 @@ class PassThroughStreamingHandler:
             and _is_message_stop_chunk(complete_frames)
             and not _is_provider_error_chunk(complete_frames)
         )
+        # An in-band ``event: error`` on a cleanly closed 200 stream is an upstream
+        # failure, not a success: log it as one (partial usage included).
+        provider_error: Final = parse_anthropic_error_event(complete_frames) if complete_frames else None
+        if endpoint_type == EndpointType.ANTHROPIC and provider_error is not None:
+            _, error_message, error_status_code = provider_error
+            await PassThroughStreamingHandler.schedule_stream_failure_logging(
+                litellm_logging_obj=litellm_logging_obj,
+                endpoint_type=endpoint_type,
+                request_body=request_body,
+                raw_bytes=raw_bytes,
+                exception=AnthropicError(status_code=error_status_code, message=error_message),
+                stream_context=PassThroughStreamContext(
+                    passthrough_success_handler_obj=passthrough_success_handler_obj,
+                    url_route=url_route,
+                    start_time=start_time,
+                ),
+            )
+            return
         try:
             # TinyFish billing is owned by the detached poller; the $0 fallback below is only for streams with no run_id
             if endpoint_type == EndpointType.TINYFISH: