diff --git a/src/finch/_base_client.py b/src/finch/_base_client.py
index b72bb8dd..6a18d451 100644
--- a/src/finch/_base_client.py
+++ b/src/finch/_base_client.py
@@ -1050,6 +1050,7 @@ def _request(
             response=response,
             stream=stream,
             stream_cls=stream_cls,
+            retries_taken=options.get_max_retries(self.max_retries) - retries,
         )
 
     def _retry_request(
@@ -1091,6 +1092,7 @@ def _process_response(
         response: httpx.Response,
         stream: bool,
         stream_cls: type[Stream[Any]] | type[AsyncStream[Any]] | None,
+        retries_taken: int = 0,
     ) -> ResponseT:
         if response.request.headers.get(RAW_RESPONSE_HEADER) == "true":
             return cast(
@@ -1102,6 +1104,7 @@ def _process_response(
                     stream=stream,
                     stream_cls=stream_cls,
                     options=options,
+                    retries_taken=retries_taken,
                 ),
             )
 
@@ -1121,6 +1124,7 @@ def _process_response(
                     stream=stream,
                     stream_cls=stream_cls,
                     options=options,
+                    retries_taken=retries_taken,
                 ),
             )
 
@@ -1134,6 +1138,7 @@ def _process_response(
             stream=stream,
             stream_cls=stream_cls,
             options=options,
+            retries_taken=retries_taken,
         )
         if bool(response.request.headers.get(RAW_RESPONSE_HEADER)):
             return cast(ResponseT, api_response)
@@ -1624,6 +1629,7 @@ async def _request(
             response=response,
             stream=stream,
             stream_cls=stream_cls,
+            retries_taken=options.get_max_retries(self.max_retries) - retries,
         )
 
     async def _retry_request(
@@ -1663,6 +1669,7 @@ async def _process_response(
         response: httpx.Response,
         stream: bool,
         stream_cls: type[Stream[Any]] | type[AsyncStream[Any]] | None,
+        retries_taken: int = 0,
     ) -> ResponseT:
         if response.request.headers.get(RAW_RESPONSE_HEADER) == "true":
             return cast(
@@ -1674,6 +1681,7 @@ async def _process_response(
                     stream=stream,
                     stream_cls=stream_cls,
                     options=options,
+                    retries_taken=retries_taken,
                 ),
             )
 
@@ -1693,6 +1701,7 @@ async def _process_response(
                     stream=stream,
                     stream_cls=stream_cls,
                     options=options,
+                    retries_taken=retries_taken,
                 ),
             )
 
@@ -1706,6 +1715,7 @@ async def _process_response(
             stream=stream,
             stream_cls=stream_cls,
             options=options,
+            retries_taken=retries_taken,
         )
         if bool(response.request.headers.get(RAW_RESPONSE_HEADER)):
             return cast(ResponseT, api_response)
diff --git a/src/finch/_legacy_response.py b/src/finch/_legacy_response.py
index c3911ba8..bba31799 100644
--- a/src/finch/_legacy_response.py
+++ b/src/finch/_legacy_response.py
@@ -5,7 +5,18 @@
 import logging
 import datetime
 import functools
-from typing import TYPE_CHECKING, Any, Union, Generic, TypeVar, Callable, Iterator, AsyncIterator, cast, overload
+from typing import (
+    TYPE_CHECKING,
+    Any,
+    Union,
+    Generic,
+    TypeVar,
+    Callable,
+    Iterator,
+    AsyncIterator,
+    cast,
+    overload,
+)
 from typing_extensions import Awaitable, ParamSpec, override, deprecated, get_origin
 
 import anyio
@@ -53,6 +64,9 @@ class LegacyAPIResponse(Generic[R]):
 
     http_response: httpx.Response
 
+    retries_taken: int
+    """The number of retries made. If no retries happened this will be `0`"""
+
     def __init__(
         self,
         *,
@@ -62,6 +76,7 @@ def __init__(
         stream: bool,
         stream_cls: type[Stream[Any]] | type[AsyncStream[Any]] | None,
         options: FinalRequestOptions,
+        retries_taken: int = 0,
     ) -> None:
         self._cast_to = cast_to
         self._client = client
@@ -70,6 +85,7 @@ def __init__(
         self._stream_cls = stream_cls
         self._options = options
         self.http_response = raw
+        self.retries_taken = retries_taken
 
     @overload
     def parse(self, *, to: type[_T]) -> _T:
diff --git a/src/finch/_response.py b/src/finch/_response.py
index df9358c9..7a0bb97a 100644
--- a/src/finch/_response.py
+++ b/src/finch/_response.py
@@ -55,6 +55,9 @@ class BaseAPIResponse(Generic[R]):
 
     http_response: httpx.Response
 
+    retries_taken: int
+    """The number of retries made. If no retries happened this will be `0`"""
+
     def __init__(
         self,
         *,
@@ -64,6 +67,7 @@ def __init__(
         stream: bool,
         stream_cls: type[Stream[Any]] | type[AsyncStream[Any]] | None,
         options: FinalRequestOptions,
+        retries_taken: int = 0,
     ) -> None:
         self._cast_to = cast_to
         self._client = client
@@ -72,6 +76,7 @@ def __init__(
         self._stream_cls = stream_cls
         self._options = options
         self.http_response = raw
+        self.retries_taken = retries_taken
 
     @property
     def headers(self) -> httpx.Headers:
diff --git a/tests/test_client.py b/tests/test_client.py
index 7863d7a1..8c43932a 100644
--- a/tests/test_client.py
+++ b/tests/test_client.py
@@ -36,6 +36,10 @@ def _get_params(client: BaseClient[Any, Any]) -> dict[str, str]:
     return dict(url.params)
 
 
+def _low_retry_timeout(*_args: Any, **_kwargs: Any) -> float:
+    return 0.1
+
+
 class TestFinch:
     client = Finch(
         base_url=base_url,
@@ -950,6 +954,49 @@ def test_parse_retry_after_header(self, remaining_retries: int, retry_after: str
         calculated = client._calculate_retry_timeout(remaining_retries, options, headers)
         assert calculated == pytest.approx(timeout, 0.5 * 0.875)  # pyright: ignore[reportUnknownMemberType]
 
+    @pytest.mark.parametrize("failures_before_success", [0, 2, 4])
+    @mock.patch("finch._base_client.BaseClient._calculate_retry_timeout", _low_retry_timeout)
+    @pytest.mark.respx(base_url=base_url)
+    def test_retries_taken(self, client: Finch, failures_before_success: int, respx_mock: MockRouter) -> None:
+        client = client.with_options(max_retries=4)
+
+        nb_retries = 0
+
+        def retry_handler(_request: httpx.Request) -> httpx.Response:
+            nonlocal nb_retries
+            if nb_retries < failures_before_success:
+                nb_retries += 1
+                return httpx.Response(500)
+            return httpx.Response(200)
+
+        respx_mock.get("/employer/directory").mock(side_effect=retry_handler)
+
+        response = client.hris.directory.with_raw_response.list()
+
+        assert response.retries_taken == failures_before_success
+
+    @pytest.mark.parametrize("failures_before_success", [0, 2, 4])
+    @mock.patch("finch._base_client.BaseClient._calculate_retry_timeout", _low_retry_timeout)
+    @pytest.mark.respx(base_url=base_url)
+    def test_retries_taken_new_response_class(
+        self, client: Finch, failures_before_success: int, respx_mock: MockRouter
+    ) -> None:
+        client = client.with_options(max_retries=4)
+
+        nb_retries = 0
+
+        def retry_handler(_request: httpx.Request) -> httpx.Response:
+            nonlocal nb_retries
+            if nb_retries < failures_before_success:
+                nb_retries += 1
+                return httpx.Response(500)
+            return httpx.Response(200)
+
+        respx_mock.get("/employer/directory").mock(side_effect=retry_handler)
+
+        with client.hris.directory.with_streaming_response.list() as response:
+            assert response.retries_taken == failures_before_success
+
 
 class TestAsyncFinch:
     client = AsyncFinch(
@@ -1870,3 +1917,50 @@ async def test_parse_retry_after_header(self, remaining_retries: int, retry_afte
         options = FinalRequestOptions(method="get", url="/foo", max_retries=3)
         calculated = client._calculate_retry_timeout(remaining_retries, options, headers)
         assert calculated == pytest.approx(timeout, 0.5 * 0.875)  # pyright: ignore[reportUnknownMemberType]
+
+    @pytest.mark.parametrize("failures_before_success", [0, 2, 4])
+    @mock.patch("finch._base_client.BaseClient._calculate_retry_timeout", _low_retry_timeout)
+    @pytest.mark.respx(base_url=base_url)
+    @pytest.mark.asyncio
+    async def test_retries_taken(
+        self, async_client: AsyncFinch, failures_before_success: int, respx_mock: MockRouter
+    ) -> None:
+        client = async_client.with_options(max_retries=4)
+
+        nb_retries = 0
+
+        def retry_handler(_request: httpx.Request) -> httpx.Response:
+            nonlocal nb_retries
+            if nb_retries < failures_before_success:
+                nb_retries += 1
+                return httpx.Response(500)
+            return httpx.Response(200)
+
+        respx_mock.get("/employer/directory").mock(side_effect=retry_handler)
+
+        response = await client.hris.directory.with_raw_response.list()
+
+        assert response.retries_taken == failures_before_success
+
+    @pytest.mark.parametrize("failures_before_success", [0, 2, 4])
+    @mock.patch("finch._base_client.BaseClient._calculate_retry_timeout", _low_retry_timeout)
+    @pytest.mark.respx(base_url=base_url)
+    @pytest.mark.asyncio
+    async def test_retries_taken_new_response_class(
+        self, async_client: AsyncFinch, failures_before_success: int, respx_mock: MockRouter
+    ) -> None:
+        client = async_client.with_options(max_retries=4)
+
+        nb_retries = 0
+
+        def retry_handler(_request: httpx.Request) -> httpx.Response:
+            nonlocal nb_retries
+            if nb_retries < failures_before_success:
+                nb_retries += 1
+                return httpx.Response(500)
+            return httpx.Response(200)
+
+        respx_mock.get("/employer/directory").mock(side_effect=retry_handler)
+
+        async with client.hris.directory.with_streaming_response.list() as response:
+            assert response.retries_taken == failures_before_success