From f70e39483c9e9a8e7e17fc6c977e58ffc87ccbb3 Mon Sep 17 00:00:00 2001
From: Otavio Brito <otaviofbrito@gmail.com>
Date: Mon, 16 Feb 2026 21:47:16 -0300
Subject: [PATCH] return actual status code - /count_tokens endpoint

---
 .../count_tokens/handler.py                   |  7 ------
 litellm/proxy/proxy_server.py                 | 24 +++++++++++++------
 2 files changed, 17 insertions(+), 14 deletions(-)

diff --git a/litellm/llms/vertex_ai/vertex_ai_partner_models/count_tokens/handler.py b/litellm/llms/vertex_ai/vertex_ai_partner_models/count_tokens/handler.py
index c6914ac3d6b..9e22f54f112 100644
--- a/litellm/llms/vertex_ai/vertex_ai_partner_models/count_tokens/handler.py
+++ b/litellm/llms/vertex_ai/vertex_ai_partner_models/count_tokens/handler.py
@@ -144,13 +144,6 @@ class VertexAIPartnerModelsTokenCounter(VertexBase):
             timeout=30.0,
         )
 
-        # Check for errors
-        if response.status_code != 200:
-            error_text = response.text
-            raise ValueError(
-                f"Token counting request failed with status {response.status_code}: {error_text}"
-            )
-
         # Parse response
         result = response.json()
 
diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py
index 294294cdda7..426561b267f 100644
--- a/litellm/proxy/proxy_server.py
+++ b/litellm/proxy/proxy_server.py
@@ -7743,13 +7743,23 @@ async def token_counter(request: TokenCountRequest, call_endpoint: bool = False)
             )
             is True
         ):
-            result = await provider_counter.count_tokens(
-                model_to_use=model_to_use or "",
-                messages=messages,  # type: ignore
-                contents=contents,
-                deployment=deployment,
-                request_model=request.model,
-            )
+            try:
+                result = await provider_counter.count_tokens(
+                    model_to_use=model_to_use or "",
+                    messages=messages,  # type: ignore
+                    contents=contents,
+                    deployment=deployment,
+                    request_model=request.model,
+                )
+            except httpx.HTTPStatusError as e:
+                error_message = getattr(e, "message", None) or str(e)
+                status_code = getattr(e, "status_code", None) or 500
+                raise ProxyException(
+                    message=error_message,
+                    type="token_counting_error",
+                    param="model",
+                    code=status_code
+                )
             #########################################################
             # Transfrom the Response to the well known format
             #########################################################