diff --git a/litellm/integrations/prometheus.py b/litellm/integrations/prometheus.py
index 2528f07f92c7..a721db5e02e3 100644
--- a/litellm/integrations/prometheus.py
+++ b/litellm/integrations/prometheus.py
@@ -10,6 +10,7 @@
 from collections.abc import Awaitable, Callable, Mapping, Sequence
 from dataclasses import replace
 from datetime import datetime, timedelta
+from functools import cache
 from typing import TYPE_CHECKING, Any, Final, Literal, Protocol, TypeAlias, TypeVar, cast
 
 from pydantic import BaseModel
@@ -151,6 +152,37 @@ def labels(self, *labelvalues: str, **labelkwargs: str) -> MetricWrapperBase:
 _MetricLike: TypeAlias = "NoOpMetric | _ExcludedLabelMetric | MetricWrapperBase"
 
 
+_ASYNC_CALL_TYPE_PREFIXES: Final = ("a_", "a")
+
+
+def _sync_twin(name: str, values_by_name: Mapping[str, str]) -> str | None:
+    return next(
+        (
+            values_by_name[name.removeprefix(prefix)]
+            for prefix in _ASYNC_CALL_TYPE_PREFIXES
+            if name.startswith(prefix) and name.removeprefix(prefix) in values_by_name
+        ),
+        None,
+    )
+
+
+@cache
+def _build_async_call_type_aliases() -> Mapping[str, str]:
+    """Matches on ``CallTypes`` member names, not values: stripping "a" from values breaks ``add_message``."""
+    from types import MappingProxyType
+
+    from litellm.types.utils import CallTypes
+
+    values_by_name: Final = MappingProxyType({member.name: str(member.value) for member in CallTypes})
+    return MappingProxyType(
+        {
+            value: twin
+            for name, value in values_by_name.items()
+            if (twin := _sync_twin(name, values_by_name)) is not None
+        }
+    )
+
+
 def _get_budget_metrics_per_request_timeout() -> float:
     raw: Final = os.getenv("PROMETHEUS_BUDGET_METRICS_PER_REQUEST_TIMEOUT")
     if raw is None:
@@ -1442,6 +1474,7 @@ async def async_log_success_event(self, kwargs, response_obj, start_time, end_ti
             model_id=standard_logging_payload["model_id"],
             api_base=standard_logging_payload["api_base"],
             api_provider=standard_logging_payload["custom_llm_provider"],
+            call_type=self._normalize_call_type(standard_logging_payload.get("call_type")),
             exception_status=None,
             exception_class=None,
             custom_metadata_labels=get_custom_labels_from_metadata(metadata=combined_metadata),
@@ -2544,6 +2577,13 @@ def _should_skip_metrics_for_invalid_key(
 
         return False
 
+    @staticmethod
+    def _normalize_call_type(call_type: str | None) -> str | None:
+        """Collapse async call types onto their sync twin so the proxy (async) and SDK (sync) share one series."""
+        if not call_type:
+            return None
+        return _build_async_call_type_aliases().get(call_type, call_type)
+
     @staticmethod
     def _extract_api_provider_from_request_data(request_data: dict) -> str | None:
         """
@@ -2639,6 +2679,9 @@ async def async_post_call_failure_hook(
                 user_agent=_metadata.get("user_agent"),
                 model_id=model_id,
                 api_provider=api_provider,
+                call_type=self._normalize_call_type(
+                    (request_data.get("standard_logging_object") or {}).get("call_type")
+                ),
                 stream=(str(request_data.get("stream")) if litellm.prometheus_emit_stream_label else None),
             )
             _label_ctx: Final = PrometheusLabelFactoryContext(enum_values)
@@ -2867,6 +2910,7 @@ def set_llm_deployment_failure_metrics(self, request_kwargs: dict):
                 tags=standard_logging_payload.get("request_tags", []),
                 client_ip=client_ip,
                 user_agent=user_agent,
+                call_type=self._normalize_call_type(standard_logging_payload.get("call_type")),
             )
 
             """
diff --git a/litellm/types/integrations/prometheus.py b/litellm/types/integrations/prometheus.py
index a024581f600e..703b4af3e1ef 100644
--- a/litellm/types/integrations/prometheus.py
+++ b/litellm/types/integrations/prometheus.py
@@ -222,6 +222,7 @@ class UserAPIKeyLabelNames(Enum):
     MCP_SERVER_NAME = "mcp_server_name"
     SERVICE_TIER = "service_tier"
     INPUT_SEQUENCE_LENGTH = "input_sequence_length"
+    CALL_TYPE = "call_type"
 
 
 DEFINED_PROMETHEUS_METRICS = Literal[
@@ -415,6 +416,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.MODEL_ID.value,
         UserAPIKeyLabelNames.API_PROVIDER.value,
         UserAPIKeyLabelNames.SERVICE_TIER.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_llm_api_time_to_first_token_metric = [
@@ -429,6 +431,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.MODEL_ID.value,
         UserAPIKeyLabelNames.API_PROVIDER.value,
         UserAPIKeyLabelNames.SERVICE_TIER.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_request_total_latency_metric = [
@@ -443,6 +446,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.MODEL_ID.value,
         UserAPIKeyLabelNames.API_PROVIDER.value,
         UserAPIKeyLabelNames.SERVICE_TIER.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_request_queue_time_seconds = [
@@ -456,6 +460,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.v1_LITELLM_MODEL_NAME.value,
         UserAPIKeyLabelNames.MODEL_ID.value,
         UserAPIKeyLabelNames.API_PROVIDER.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     # Guardrail metrics - these use custom labels (guardrail_name, status, error_type, hook_type)
@@ -479,6 +484,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.USER_AGENT.value,
         UserAPIKeyLabelNames.MODEL_ID.value,
         UserAPIKeyLabelNames.API_PROVIDER.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_proxy_failed_requests_metric = [
@@ -501,6 +507,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.USER_AGENT.value,
         UserAPIKeyLabelNames.MODEL_ID.value,
         UserAPIKeyLabelNames.API_PROVIDER.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_deployment_latency_per_output_token = [
@@ -512,6 +519,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.API_KEY_ALIAS.value,
         UserAPIKeyLabelNames.TEAM.value,
         UserAPIKeyLabelNames.TEAM_ALIAS.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_overhead_latency_metric = [
@@ -522,6 +530,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.API_KEY_HASH.value,
         UserAPIKeyLabelNames.API_KEY_ALIAS.value,
         UserAPIKeyLabelNames.MODEL_ID.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_overhead_with_guardrails_latency_metric = [
@@ -532,6 +541,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.API_KEY_HASH.value,
         UserAPIKeyLabelNames.API_KEY_ALIAS.value,
         UserAPIKeyLabelNames.MODEL_ID.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_remaining_requests_metric = [
@@ -568,6 +578,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.REQUESTED_MODEL.value,
         UserAPIKeyLabelNames.MODEL_ID.value,
         UserAPIKeyLabelNames.API_PROVIDER.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_spend_metric = [
@@ -585,6 +596,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.MODEL_ID.value,
         UserAPIKeyLabelNames.API_PROVIDER.value,
         UserAPIKeyLabelNames.SERVICE_TIER.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_input_tokens_metric = [
@@ -599,6 +611,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.REQUESTED_MODEL.value,
         UserAPIKeyLabelNames.MODEL_ID.value,
         UserAPIKeyLabelNames.API_PROVIDER.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_total_tokens_metric = [
@@ -613,6 +626,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.REQUESTED_MODEL.value,
         UserAPIKeyLabelNames.MODEL_ID.value,
         UserAPIKeyLabelNames.API_PROVIDER.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_output_tokens_metric = [
@@ -627,6 +641,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.REQUESTED_MODEL.value,
         UserAPIKeyLabelNames.MODEL_ID.value,
         UserAPIKeyLabelNames.API_PROVIDER.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     # Token-type detail metrics — reuse the same label set as
@@ -764,6 +779,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.TEAM_ALIAS.value,
         UserAPIKeyLabelNames.CLIENT_IP.value,
         UserAPIKeyLabelNames.USER_AGENT.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_deployment_total_requests = [
@@ -778,6 +794,7 @@ class PrometheusMetricLabels:
         UserAPIKeyLabelNames.TEAM_ALIAS.value,
         UserAPIKeyLabelNames.CLIENT_IP.value,
         UserAPIKeyLabelNames.USER_AGENT.value,
+        UserAPIKeyLabelNames.CALL_TYPE.value,
     ]
 
     litellm_deployment_success_responses = litellm_deployment_total_requests
@@ -1033,6 +1050,7 @@ class UserAPIKeyLabelValues:
     model_id: str | None = None
     api_base: str | None = None
     api_provider: str | None = None
+    call_type: str | None = None
     exception_status: str | None = None
     exception_class: str | None = None
     rate_limit_category: str | None = None
diff --git a/tests/enterprise/litellm_enterprise/enterprise_callbacks/test_prometheus_logging_callbacks.py b/tests/enterprise/litellm_enterprise/enterprise_callbacks/test_prometheus_logging_callbacks.py
index 58cde4c8103e..d8520d20378b 100644
--- a/tests/enterprise/litellm_enterprise/enterprise_callbacks/test_prometheus_logging_callbacks.py
+++ b/tests/enterprise/litellm_enterprise/enterprise_callbacks/test_prometheus_logging_callbacks.py
@@ -244,6 +244,7 @@ def test_increment_token_metrics(prometheus_logger):
         model="gpt-5-mini",
         model_id="model-123",
         api_provider="openai",
+        call_type="completion",
     )
     prometheus_logger.litellm_tokens_metric.labels().inc.assert_called_once_with(100)
 
@@ -261,6 +262,7 @@ def test_increment_token_metrics(prometheus_logger):
         model="gpt-5-mini",
         model_id="model-123",
         api_provider="openai",
+        call_type="completion",
     )
     prometheus_logger.litellm_input_tokens_metric.labels().inc.assert_called_once_with(
         50
@@ -280,6 +282,7 @@ def test_increment_token_metrics(prometheus_logger):
         model="gpt-5-mini",
         model_id="model-123",
         api_provider="openai",
+        call_type="completion",
     )
     prometheus_logger.litellm_output_tokens_metric.labels().inc.assert_called_once_with(
         50
@@ -443,6 +446,7 @@ def test_set_latency_metrics(prometheus_logger):
         model="gpt-5-mini",
         model_id="model-123",
         api_provider="openai",
+        call_type="completion",
         service_tier=None,
     )
     prometheus_logger.litellm_llm_api_time_to_first_token_metric.labels().observe.assert_called_once_with(
@@ -463,6 +467,7 @@ def test_set_latency_metrics(prometheus_logger):
         model="gpt-5-mini",
         model_id="model-123",
         api_provider="openai",
+        call_type="completion",
         service_tier=None,
     )
     prometheus_logger.litellm_llm_api_latency_metric.labels().observe.assert_called_once_with(
@@ -483,6 +488,7 @@ def test_set_latency_metrics(prometheus_logger):
         model="gpt-5-mini",
         model_id="model-123",
         api_provider="openai",
+        call_type="completion",
         service_tier=None,
     )
     prometheus_logger.litellm_request_total_latency_metric.labels().observe.assert_called_once_with(
@@ -629,6 +635,7 @@ def test_increment_top_level_request_and_spend_metrics(prometheus_logger):
         model="gpt-5-mini",
         model_id="model-123",
         api_provider="openai",
+        call_type="completion",
         client_ip=None,
         user_agent=None,
         requested_model=None,
@@ -649,6 +656,7 @@ def test_increment_top_level_request_and_spend_metrics(prometheus_logger):
         model="gpt-5-mini",
         model_id="model-123",
         api_provider="openai",
+        call_type="completion",
         client_ip=None,
         user_agent=None,
         requested_model=None,
@@ -870,6 +878,7 @@ async def test_async_post_call_failure_hook(prometheus_logger, known_model_route
             client_ip=None,
             user_agent=None,
             api_provider="openai",
+            call_type=None,
         )
     finally:
         litellm.prometheus_emit_rate_limit_labels = original_emit
@@ -894,6 +903,7 @@ async def test_async_post_call_failure_hook(prometheus_logger, known_model_route
         client_ip=None,
         user_agent=None,
         api_provider="openai",
+        call_type=None,
     )
     prometheus_logger.litellm_proxy_total_requests_metric.labels().inc.assert_called_once()
 
@@ -1030,6 +1040,7 @@ def test_set_llm_deployment_success_metrics(prometheus_logger):
         model_id="model-123",
         api_base="https://api.openai.com",
         api_provider="openai",
+        call_type="completion",
         requested_model="my_custom_model_group",
         hashed_api_key=standard_logging_payload["metadata"]["user_api_key_hash"],
         api_key_alias=standard_logging_payload["metadata"]["user_api_key_alias"],
@@ -1046,6 +1057,7 @@ def test_set_llm_deployment_success_metrics(prometheus_logger):
         model_id="model-123",
         api_base="https://api.openai.com",
         api_provider="openai",
+        call_type="completion",
         requested_model="my_custom_model_group",
         hashed_api_key=standard_logging_payload["metadata"]["user_api_key_hash"],
         api_key_alias=standard_logging_payload["metadata"]["user_api_key_alias"],
@@ -1062,6 +1074,7 @@ def test_set_llm_deployment_success_metrics(prometheus_logger):
         model_id="model-123",
         api_base="https://api.openai.com",
         api_provider="openai",
+        call_type="completion",
         hashed_api_key=standard_logging_payload["metadata"]["user_api_key_hash"],
         api_key_alias=standard_logging_payload["metadata"]["user_api_key_alias"],
         team=standard_logging_payload["metadata"]["user_api_key_team_id"],
@@ -1073,6 +1086,7 @@ def test_set_llm_deployment_success_metrics(prometheus_logger):
         api_base="https://api.openai.com",
         api_key_alias=standard_logging_payload["metadata"]["user_api_key_alias"],
         api_provider="openai",
+        call_type="completion",
         hashed_api_key=standard_logging_payload["metadata"]["user_api_key_hash"],
         litellm_model_name="gpt-5-mini",
         model_group="my_custom_model_group",
diff --git a/tests/test_litellm/integrations/test_prometheus_labels.py b/tests/test_litellm/integrations/test_prometheus_labels.py
index 859cdd30c117..8d3a98d15192 100644
--- a/tests/test_litellm/integrations/test_prometheus_labels.py
+++ b/tests/test_litellm/integrations/test_prometheus_labels.py
@@ -629,7 +629,7 @@ async def test_success_hook_emits_api_provider_value_on_token_metric():
 
     payload = {
         "id": "t",
-        "call_type": "completion",
+        "call_type": "acompletion",
         "response_cost": 0.001,
         "status": "success",
         "total_tokens": 30,
@@ -677,6 +677,7 @@ async def test_success_hook_emits_api_provider_value_on_token_metric():
         )
         samples = _collected_samples("litellm_total_tokens_metric_total")
         assert samples, "expected litellm_total_tokens_metric to be emitted"
+        assert all(s.labels.get("call_type") == "completion" for s in samples)
         assert all(s.labels.get("api_provider") == "openai" for s in samples), (
             "collected token metric must carry api_provider=openai, got "
             f"{[s.labels.get('api_provider') for s in samples]}"
@@ -702,12 +703,17 @@ async def test_failure_hook_emits_api_provider_value_on_failed_requests_metric()
     try:
         logger = PrometheusLogger()
         await logger.async_post_call_failure_hook(
-            request_data={"model": "gpt-4o-mini", "metadata": {}},
+            request_data={
+                "model": "gpt-4o-mini",
+                "metadata": {},
+                "standard_logging_object": {"call_type": "acompletion"},
+            },
             original_exception=Exception("boom"),
             user_api_key_dict=UserAPIKeyAuth(token="tok"),
         )
         samples = _collected_samples("litellm_proxy_failed_requests_metric_total")
         assert samples, "expected litellm_proxy_failed_requests_metric to be emitted"
+        assert all(s.labels.get("call_type") == "completion" for s in samples)
         assert any(s.labels.get("api_provider") == "openai" for s in samples), (
             "collected failed-requests metric must carry api_provider=openai, got "
             f"{[s.labels.get('api_provider') for s in samples]}"
@@ -730,3 +736,116 @@ async def test_failure_hook_emits_api_provider_value_on_failed_requests_metric()
     test_prometheus_label_value_sanitization_none()
     test_prometheus_label_value_sanitization_non_string_types()
     print("\n✅ All prometheus label tests passed!")
+
+
+def test_call_type_in_request_lifecycle_metrics():
+    call_type_label = UserAPIKeyLabelNames.CALL_TYPE.value
+
+    metrics_with_call_type = [
+        "litellm_spend_metric",
+        "litellm_requests_metric",
+        "litellm_input_tokens_metric",
+        "litellm_output_tokens_metric",
+        "litellm_total_tokens_metric",
+        "litellm_llm_api_latency_metric",
+        "litellm_llm_api_time_to_first_token_metric",
+        "litellm_request_total_latency_metric",
+        "litellm_request_queue_time_seconds",
+        "litellm_overhead_latency_metric",
+        "litellm_overhead_with_guardrails_latency_metric",
+        "litellm_proxy_total_requests_metric",
+        "litellm_proxy_failed_requests_metric",
+        "litellm_deployment_latency_per_output_token",
+        "litellm_deployment_failure_responses",
+        "litellm_deployment_total_requests",
+        "litellm_deployment_success_responses",
+    ]
+
+    for metric_name in metrics_with_call_type:
+        labels = PrometheusMetricLabels.get_labels(metric_name)
+        assert call_type_label in labels, f"Metric {metric_name} should contain call_type"
+
+
+def test_call_type_absent_from_quota_and_deployment_gauges():
+    """Headroom belongs to a deployment; splitting it by call type would emit duplicate series for one limit."""
+    call_type_label = UserAPIKeyLabelNames.CALL_TYPE.value
+
+    for metric_name in [
+        "litellm_remaining_requests_metric",
+        "litellm_remaining_tokens_metric",
+        "litellm_deployment_tpm_limit",
+        "litellm_deployment_rpm_limit",
+        "litellm_deployment_state",
+        "litellm_deployment_cooled_down",
+    ]:
+        labels = PrometheusMetricLabels.get_labels(metric_name)
+        assert call_type_label not in labels, f"Metric {metric_name} should not carry call_type"
+
+
+@pytest.mark.parametrize(
+    "raw, expected",
+    [
+        ("acompletion", "completion"),
+        ("completion", "completion"),
+        ("aembedding", "embedding"),
+        ("aimage_generation", "image_generation"),
+        ("a_add_message", "add_message"),
+        ("aanthropic_messages", "anthropic_messages"),
+        ("add_message", "add_message"),
+        ("anthropic_messages", "anthropic_messages"),
+        ("something_new", "something_new"),
+        (None, None),
+        ("", None),
+    ],
+)
+def test_normalize_call_type(raw, expected):
+    from litellm.integrations.prometheus import PrometheusLogger
+
+    assert PrometheusLogger._normalize_call_type(raw) == expected
+
+
+def test_async_call_type_aliases_cover_every_async_member():
+    from litellm.integrations.prometheus import PrometheusLogger
+    from litellm.types.utils import CallTypes
+
+    names = {member.name: member.value for member in CallTypes}
+    for name, value in names.items():
+        for prefix in ("a_", "a"):
+            twin = name[len(prefix) :] if name.startswith(prefix) else None
+            if twin and twin in names and twin != name:
+                assert PrometheusLogger._normalize_call_type(value) == names[twin]
+                break
+
+
+def test_call_type_reaches_rendered_metrics_output():
+    from litellm.integrations.prometheus import PrometheusLogger
+
+    _clear_prometheus_registry()
+    logger = PrometheusLogger()
+
+    standard_logging_payload = {
+        "model_group": "gpt-4o-group",
+        "call_type": "acompletion",
+        "api_base": "https://example.invalid",
+        "model_id": "deployment-1",
+        "request_tags": [],
+        "metadata": {
+            "user_api_key_hash": "test-key",
+            "user_api_key_alias": None,
+            "user_api_key_team_id": None,
+            "user_api_key_team_alias": None,
+        },
+    }
+
+    logger.set_llm_deployment_failure_metrics(
+        {
+            "model": "gpt-4o",
+            "litellm_params": {"custom_llm_provider": "openai"},
+            "standard_logging_object": standard_logging_payload,
+            "exception": Exception("boom"),
+        }
+    )
+
+    samples = _collected_samples("litellm_deployment_failure_responses_total")
+    assert samples, "expected litellm_deployment_failure_responses to be emitted"
+    assert samples[0].labels.get("call_type") == "completion"