diff --git a/litellm/__pycache__/utils.cpython-311.pyc b/litellm/__pycache__/utils.cpython-311.pyc
index 27a856762..5ef3ac7b1 100644
Binary files a/litellm/__pycache__/utils.cpython-311.pyc and b/litellm/__pycache__/utils.cpython-311.pyc differ
diff --git a/litellm/llms/anthropic.py b/litellm/llms/anthropic.py
index c4e2ece7c..904c81819 100644
--- a/litellm/llms/anthropic.py
+++ b/litellm/llms/anthropic.py
@@ -113,10 +113,13 @@ class AnthropicLLM:
             )
             print_verbose(f"raw model_response: {response.text}")
             ## RESPONSE OBJECT
-            completion_response = response.json()
+            try:
+                completion_response = response.json()
+            except:
+                raise AnthropicError(message=response.text, status_code=response.status_code)
             if "error" in completion_response:
                 raise AnthropicError(
-                    message=completion_response["error"],
+                    message=str(completion_response["error"]),
                     status_code=response.status_code,
                 )
             else:
diff --git a/litellm/tests/test_exceptions.py b/litellm/tests/test_exceptions.py
index 819aade14..80376ed3d 100644
--- a/litellm/tests/test_exceptions.py
+++ b/litellm/tests/test_exceptions.py
@@ -23,7 +23,6 @@ litellm.vertex_project = "pathrise-convert-1606954137718"
 litellm.vertex_location = "us-central1"
 
 litellm.failure_callback = ["sentry"]
-# litellm.set_verbose = True
 #### What this tests ####
 #    This tests exception mapping -> trigger an exception from an llm provider -> assert if output is of the expected type
 
@@ -35,8 +34,8 @@ litellm.failure_callback = ["sentry"]
 # Approach: Run each model through the test -> assert if the correct error (always the same one) is triggered
 
 # models = ["gpt-3.5-turbo", "chatgpt-test",  "claude-instant-1", "command-nightly"]
-test_model = "gpt-3.5-turbo"
-models = ["gpt-3.5-turbo"]
+test_model = "claude-instant-1"
+models = ["claude-instant-1"]
 
 
 def logging_fn(model_call_dict):
@@ -50,7 +49,7 @@ def logging_fn(model_call_dict):
 # Test 1: Context Window Errors
 @pytest.mark.parametrize("model", models)
 def test_context_window(model):
-    sample_text = "how does a court case get to the Supreme Court?" * 1000000
+    sample_text = "how does a court case get to the Supreme Court?" * 50000
     messages = [{"content": sample_text, "role": "user"}]
     try:
         print(f"model: {model}")
@@ -64,6 +63,7 @@ def test_context_window(model):
         return
     except InvalidRequestError as e:
         print(f"InvalidRequestError: {e.llm_provider}")
+        print(f"InvalidRequestError message: {e.message}")
         return
     except OpenAIError as e:
         print(f"OpenAIError: {e.llm_provider}")
diff --git a/litellm/utils.py b/litellm/utils.py
index a43da00c8..f802e7f69 100644
--- a/litellm/utils.py
+++ b/litellm/utils.py
@@ -1357,6 +1357,14 @@ def exception_type(model, original_exception, custom_llm_provider):
             else:
                 exception_type = ""
             if "claude" in model:  # one of the anthropics
+                if hasattr(original_exception, "message"):
+                    if "prompt is too long" in original_exception.message:
+                        exception_mapping_worked = True
+                        raise ContextWindowExceededError(
+                            message=original_exception.message, 
+                            model=model,
+                            llm_provider="anthropic"
+                        )
                 if hasattr(original_exception, "status_code"):
                     print_verbose(f"status_code: {original_exception.status_code}")
                     if original_exception.status_code == 401:
@@ -1372,6 +1380,12 @@ def exception_type(model, original_exception, custom_llm_provider):
                             model=model,
                             llm_provider="anthropic",
                         )
+                    elif original_exception.status_code == 413:
+                        exception_mapping_worked = True
+                        raise ContextWindowExceededError(
+                            message=f"AnthropicException - {original_exception.message}",
+                            llm_provider="anthropic",
+                        )
                     elif original_exception.status_code == 429:
                         exception_mapping_worked = True
                         raise RateLimitError(