diff --git a/litellm/proxy/pass_through_endpoints/llm_provider_handlers/azure_speech_passthrough_logging_handler.py b/litellm/proxy/pass_through_endpoints/llm_provider_handlers/azure_speech_passthrough_logging_handler.py
index 23215fb8978..8fb5b88af1c 100644
--- a/litellm/proxy/pass_through_endpoints/llm_provider_handlers/azure_speech_passthrough_logging_handler.py
+++ b/litellm/proxy/pass_through_endpoints/llm_provider_handlers/azure_speech_passthrough_logging_handler.py
@@ -40,7 +40,7 @@ class _EnhancedMode(BaseModel):
 class _FastTranscriptionDefinition(BaseModel):
     model_config = ConfigDict(frozen=True, strict=True)
 
-    enhancedMode: _EnhancedMode | None = None
+    enhanced_mode: _EnhancedMode | None = Field(default=None, alias="enhancedMode")
 
 
 class AzureSpeechPassthroughLoggingHandler:
@@ -55,9 +55,20 @@ class AzureSpeechPassthroughLoggingHandler:
             parsed: Final = _FastTranscriptionDefinition.model_validate_json(definition)
         except ValidationError:
             return None
-        if parsed.enhancedMode is None or not parsed.enhancedMode.enabled:
+        if parsed.enhanced_mode is None or not parsed.enhanced_mode.enabled:
             return None
-        return parsed.enhancedMode.model
+        return parsed.enhanced_mode.model
+
+    @staticmethod
+    def _resolve_model_and_pricing_model(url_route: str, request_body: Mapping[str, object]) -> tuple[str, str]:
+        enhanced_model: Final = AzureSpeechPassthroughLoggingHandler._enhanced_transcription_model(
+            url_route, request_body
+        )
+        if enhanced_model is None:
+            return AzureSpeechPassthroughLoggingHandler._model_from_url_route(url_route), AZURE_SPEECH_PRICING_MODEL
+        model_name: Final = f"{AZURE_SPEECH_CUSTOM_LLM_PROVIDER}/{enhanced_model}"
+        pricing_model: Final = get_model_cost_key(f"azure/speech/{enhanced_model}")
+        return model_name, pricing_model or AZURE_SPEECH_PRICING_MODEL
 
     @staticmethod
     def _is_short_audio_route(url_route: str) -> bool:
@@ -163,18 +174,9 @@ class AzureSpeechPassthroughLoggingHandler:
         **kwargs: object,  # kwargs-ok: the passthrough logging dispatch forwards shared logging kwargs to every handler
     ) -> PassThroughEndpointLoggingTypedDict:
         try:
-            enhanced_model: Final = AzureSpeechPassthroughLoggingHandler._enhanced_transcription_model(
+            model_name, pricing_model = AzureSpeechPassthroughLoggingHandler._resolve_model_and_pricing_model(
                 url_route, request_body
             )
-            model_name: Final = (
-                f"{AZURE_SPEECH_CUSTOM_LLM_PROVIDER}/{enhanced_model}"
-                if enhanced_model is not None
-                else AzureSpeechPassthroughLoggingHandler._model_from_url_route(url_route)
-            )
-            enhanced_pricing_model: Final = (
-                get_model_cost_key(f"azure/speech/{enhanced_model}") if enhanced_model is not None else None
-            )
-            pricing_model: Final = enhanced_pricing_model or AZURE_SPEECH_PRICING_MODEL
             response_cost: Final = AzureSpeechPassthroughLoggingHandler._response_cost(
                 url_route, httpx_response, response_body, pricing_model
             )
diff --git a/tests/unit/proxy/pass_through_endpoints/test_llm_pass_through_endpoints.py b/tests/unit/proxy/pass_through_endpoints/test_llm_pass_through_endpoints.py
index f3b827191a9..aa6ca6a78ef 100644
--- a/tests/unit/proxy/pass_through_endpoints/test_llm_pass_through_endpoints.py
+++ b/tests/unit/proxy/pass_through_endpoints/test_llm_pass_through_endpoints.py
@@ -7700,6 +7700,13 @@ class TestAzureSpeechProxyRoute:
     def test_fast_transcription_multipart_upload_is_forwarded_byte_for_byte(
         self, azure_speech_client: TestClient
     ) -> None:
+        upload: Final = httpx.Request(
+            "POST",
+            f"http://testserver/azure_speech{AZURE_SPEECH_FAST_ENDPOINT}",
+            files={"audio": ("eagle.wav", AZURE_SPEECH_NON_UTF8_WAV_BYTES, "audio/wav")},
+            data={"definition": json.dumps({"locales": ["en-US"]})},
+        )
+        uploaded_body: Final = upload.read()
         with respx.mock(assert_all_called=True) as upstream:
             route = upstream.post(f"https://eastus.api.cognitive.microsoft.com{AZURE_SPEECH_FAST_ENDPOINT}").mock(
                 return_value=httpx.Response(200, json={"durationMilliseconds": 640, "combinedPhrases": []})
@@ -7708,17 +7715,18 @@ class TestAzureSpeechProxyRoute:
             response = azure_speech_client.post(
                 f"/azure_speech{AZURE_SPEECH_FAST_ENDPOINT}",
                 params={"api-version": "2024-11-15"},
-                files={"audio": ("eagle.wav", AZURE_SPEECH_NON_UTF8_WAV_BYTES, "audio/wav")},
-                data={"definition": json.dumps({"locales": ["en-US"]})},
-                headers={"Authorization": "Bearer sk-virtual"},
+                content=uploaded_body,
+                headers={
+                    "Authorization": "Bearer sk-virtual",
+                    "Content-Type": upload.headers["content-type"],
+                },
             )
 
         assert response.status_code == 200
         sent = route.calls.last.request
-        assert sent.headers["content-type"].startswith("multipart/form-data; boundary=")
+        assert sent.headers["content-type"] == upload.headers["content-type"]
         assert dict(sent.url.params) == {"api-version": "2024-11-15"}
-        assert AZURE_SPEECH_NON_UTF8_WAV_BYTES in sent.content
-        assert b'name="definition"' in sent.content
+        assert sent.content == uploaded_body
         assert sent.headers["ocp-apim-subscription-key"] == "server-subscription-key"
         assert "authorization" not in sent.headers
 
@@ -7769,9 +7777,21 @@ class TestAzureSpeechProxyRoute:
             ("azure_speech/batch-transcription", "azure_speech", 0.0)
         ]
 
-    @pytest.mark.parametrize("enhanced_model", [None, "MAI-Transcribe-2", "MAI-Transcribe-1.5", "future-model"])
+    @pytest.mark.parametrize(
+        "enhanced_model,pricing_model",
+        [
+            (None, "azure/speech/azure-stt"),
+            ("MAI-Transcribe-2", "azure/speech/MAI-Transcribe-2"),
+            ("MAI-Transcribe-1.5", "azure/speech/MAI-Transcribe-1.5"),
+            ("future-model", "azure/speech/azure-stt"),
+        ],
+    )
     def test_fast_transcription_spend_is_priced_from_duration_milliseconds(
-        self, azure_speech_client: TestClient, monkeypatch: pytest.MonkeyPatch, enhanced_model: str | None
+        self,
+        azure_speech_client: TestClient,
+        monkeypatch: pytest.MonkeyPatch,
+        enhanced_model: str | None,
+        pricing_model: str,
     ) -> None:
         from litellm.integrations.custom_logger import CustomLogger
 
@@ -7795,8 +7815,7 @@ class TestAzureSpeechProxyRoute:
                 "output_cost_per_second": 0.0,
             },
         )
-        selected_price: Final = litellm.model_cost.get(f"azure/speech/{enhanced_model}")
-        expected_rate: Final = selected_price["input_cost_per_second"] if selected_price is not None else 0.25
+        expected_rate: Final = litellm.model_cost[pricing_model]["input_cost_per_second"]
         definition: Final = json.dumps(
             {"locales": ["en-US"], "enhancedMode": {"enabled": True, "model": enhanced_model}}
             if enhanced_model is not None