diff --git a/include/ur_api.h b/include/ur_api.h
index 076000308f..bdaee31486 100644
--- a/include/ur_api.h
+++ b/include/ur_api.h
@@ -9486,13 +9486,17 @@ urEnqueueCooperativeKernelLaunchExp(
 ///     - ::UR_RESULT_ERROR_INVALID_NULL_HANDLE
 ///         + `NULL == hKernel`
 ///     - ::UR_RESULT_ERROR_INVALID_NULL_POINTER
+///         + `NULL == pLocalWorkSize`
 ///         + `NULL == pGroupCountRet`
 ///     - ::UR_RESULT_ERROR_INVALID_KERNEL
 UR_APIEXPORT ur_result_t UR_APICALL
 urKernelSuggestMaxCooperativeGroupCountExp(
     ur_kernel_handle_t hKernel,     ///< [in] handle of the kernel object
-    size_t localWorkSize,           ///< [in] number of local work-items that will form a work-group when the
-                                    ///< kernel is launched
+    uint32_t workDim,               ///< [in] number of dimensions, from 1 to 3, to specify the work-group
+                                    ///< work-items
+    const size_t *pLocalWorkSize,   ///< [in] pointer to an array of workDim unsigned values that specify the
+                                    ///< number of local work-items forming a work-group that will execute the
+                                    ///< kernel function.
     size_t dynamicSharedMemorySize, ///< [in] size of dynamic shared memory, for each work-group, in bytes,
                                     ///< that will be used when the kernel is launched
     uint32_t *pGroupCountRet        ///< [out] pointer to maximum number of groups
@@ -11028,7 +11032,8 @@ typedef struct ur_kernel_set_specialization_constants_params_t {
 ///     allowing the callback the ability to modify the parameter's value
 typedef struct ur_kernel_suggest_max_cooperative_group_count_exp_params_t {
     ur_kernel_handle_t *phKernel;
-    size_t *plocalWorkSize;
+    uint32_t *pworkDim;
+    const size_t **ppLocalWorkSize;
     size_t *pdynamicSharedMemorySize;
     uint32_t **ppGroupCountRet;
 } ur_kernel_suggest_max_cooperative_group_count_exp_params_t;
diff --git a/include/ur_ddi.h b/include/ur_ddi.h
index d4b2595d7f..ce7dd137a9 100644
--- a/include/ur_ddi.h
+++ b/include/ur_ddi.h
@@ -651,7 +651,8 @@ typedef ur_result_t(UR_APICALL *ur_pfnGetKernelProcAddrTable_t)(
 /// @brief Function-pointer for urKernelSuggestMaxCooperativeGroupCountExp
 typedef ur_result_t(UR_APICALL *ur_pfnKernelSuggestMaxCooperativeGroupCountExp_t)(
     ur_kernel_handle_t,
-    size_t,
+    uint32_t,
+    const size_t *,
     size_t,
     uint32_t *);
 
diff --git a/include/ur_print.hpp b/include/ur_print.hpp
index 3aa165fcd0..8284731dc1 100644
--- a/include/ur_print.hpp
+++ b/include/ur_print.hpp
@@ -13074,9 +13074,15 @@ inline std::ostream &operator<<(std::ostream &os, [[maybe_unused]] const struct
                           *(params->phKernel));
 
     os << ", ";
-    os << ".localWorkSize = ";
+    os << ".workDim = ";
+
+    os << *(params->pworkDim);
+
+    os << ", ";
+    os << ".pLocalWorkSize = ";
 
-    os << *(params->plocalWorkSize);
+    ur::details::printPtr(os,
+                          *(params->ppLocalWorkSize));
 
     os << ", ";
     os << ".dynamicSharedMemorySize = ";
diff --git a/scripts/core/exp-cooperative-kernels.yml b/scripts/core/exp-cooperative-kernels.yml
index 941aba29fa..ad3ba0ffba 100644
--- a/scripts/core/exp-cooperative-kernels.yml
+++ b/scripts/core/exp-cooperative-kernels.yml
@@ -78,9 +78,13 @@ params:
     - type: $x_kernel_handle_t
       name: hKernel
       desc: "[in] handle of the kernel object"
-    - type: size_t
-      name: localWorkSize
-      desc: "[in] number of local work-items that will form a work-group when the kernel is launched"
+    - type: uint32_t
+      name: workDim
+      desc: "[in] number of dimensions, from 1 to 3, to specify the work-group work-items"
+    - type: "const size_t*"
+      name: pLocalWorkSize
+      desc: |
+            [in] pointer to an array of workDim unsigned values that specify the number of local work-items forming a work-group that will execute the kernel function.
     - type: size_t
       name: dynamicSharedMemorySize
       desc: "[in] size of dynamic shared memory, for each work-group, in bytes, that will be used when the kernel is launched"
diff --git a/source/adapters/cuda/kernel.cpp b/source/adapters/cuda/kernel.cpp
index 5fb097c304..46c4907d4b 100644
--- a/source/adapters/cuda/kernel.cpp
+++ b/source/adapters/cuda/kernel.cpp
@@ -190,10 +190,14 @@ UR_APIEXPORT ur_result_t UR_APICALL urKernelGetNativeHandle(
 }
 
 UR_APIEXPORT ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
-    ur_kernel_handle_t hKernel, size_t localWorkSize,
+    ur_kernel_handle_t hKernel, uint32_t workDim, const size_t *pLocalWorkSize,
     size_t dynamicSharedMemorySize, uint32_t *pGroupCountRet) {
   UR_ASSERT(hKernel, UR_RESULT_ERROR_INVALID_KERNEL);
 
+  size_t localWorkSize = pLocalWorkSize[0];
+  localWorkSize *= (workDim >= 2 ? pLocalWorkSize[1] : 1);
+  localWorkSize *= (workDim == 3 ? pLocalWorkSize[2] : 1);
+
   // We need to set the active current device for this kernel explicitly here,
   // because the occupancy querying API does not take device parameter.
   ur_device_handle_t Device = hKernel->getProgram()->getDevice();
diff --git a/source/adapters/hip/kernel.cpp b/source/adapters/hip/kernel.cpp
index 60931cd014..176a2a495a 100644
--- a/source/adapters/hip/kernel.cpp
+++ b/source/adapters/hip/kernel.cpp
@@ -169,10 +169,11 @@ urKernelGetNativeHandle(ur_kernel_handle_t, ur_native_handle_t *) {
 }
 
 UR_APIEXPORT ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
-    ur_kernel_handle_t hKernel, size_t localWorkSize,
+    ur_kernel_handle_t hKernel, uint32_t workDim, const size_t *pLocalWorkSize,
     size_t dynamicSharedMemorySize, uint32_t *pGroupCountRet) {
   std::ignore = hKernel;
-  std::ignore = localWorkSize;
+  std::ignore = workDim;
+  std::ignore = pLocalWorkSize;
   std::ignore = dynamicSharedMemorySize;
   std::ignore = pGroupCountRet;
   return UR_RESULT_ERROR_UNSUPPORTED_FEATURE;
diff --git a/source/adapters/level_zero/kernel.cpp b/source/adapters/level_zero/kernel.cpp
index 153d3861b1..b15b4ce147 100644
--- a/source/adapters/level_zero/kernel.cpp
+++ b/source/adapters/level_zero/kernel.cpp
@@ -1054,11 +1054,17 @@ ur_result_t urKernelGetNativeHandle(
 }
 
 ur_result_t urKernelSuggestMaxCooperativeGroupCountExp(
-    ur_kernel_handle_t hKernel, size_t localWorkSize,
+    ur_kernel_handle_t hKernel, uint32_t workDim, const size_t *pLocalWorkSize,
     size_t dynamicSharedMemorySize, uint32_t *pGroupCountRet) {
-  (void)localWorkSize;
   (void)dynamicSharedMemorySize;
   std::shared_lock<ur_shared_mutex> Guard(hKernel->Mutex);
+
+  uint32_t WG[3];
+  WG[0] = ur_cast<uint32_t>(pLocalWorkSize[0]);
+  WG[1] = workDim >= 2 ? ur_cast<uint32_t>(pLocalWorkSize[1]) : 1;
+  WG[2] = workDim == 3 ? ur_cast<uint32_t>(pLocalWorkSize[2]) : 1;
+  ZE2UR_CALL(zeKernelSetGroupSize, (hKernel->ZeKernel, WG[0], WG[1], WG[2]));
+
   uint32_t TotalGroupCount = 0;
   ZE2UR_CALL(zeKernelSuggestMaxCooperativeGroupCount,
              (hKernel->ZeKernel, &TotalGroupCount));
diff --git a/source/adapters/level_zero/ur_interface_loader.hpp b/source/adapters/level_zero/ur_interface_loader.hpp
index 6097d8ffd8..ed7e97fa3b 100644
--- a/source/adapters/level_zero/ur_interface_loader.hpp
+++ b/source/adapters/level_zero/ur_interface_loader.hpp
@@ -687,7 +687,7 @@ ur_result_t urEnqueueCooperativeKernelLaunchExp(
     const size_t *pLocalWorkSize, uint32_t numEventsInWaitList,
     const ur_event_handle_t *phEventWaitList, ur_event_handle_t *phEvent);
 ur_result_t urKernelSuggestMaxCooperativeGroupCountExp(
-    ur_kernel_handle_t hKernel, size_t localWorkSize,
+    ur_kernel_handle_t hKernel, uint32_t workDim, const size_t *pLocalWorkSize,
     size_t dynamicSharedMemorySize, uint32_t *pGroupCountRet);
 ur_result_t urEnqueueTimestampRecordingExp(
     ur_queue_handle_t hQueue, bool blocking, uint32_t numEventsInWaitList,
diff --git a/source/adapters/level_zero/v2/api.cpp b/source/adapters/level_zero/v2/api.cpp
index e4a70df811..593115a99f 100644
--- a/source/adapters/level_zero/v2/api.cpp
+++ b/source/adapters/level_zero/v2/api.cpp
@@ -568,7 +568,7 @@ ur_result_t urCommandBufferCommandGetInfoExp(
 }
 
 ur_result_t urKernelSuggestMaxCooperativeGroupCountExp(
-    ur_kernel_handle_t hKernel, size_t localWorkSize,
+    ur_kernel_handle_t hKernel, uint32_t workDim, const size_t *pLocalWorkSize,
     size_t dynamicSharedMemorySize, uint32_t *pGroupCountRet) {
   logger::error("{} function not implemented!", __FUNCTION__);
   return UR_RESULT_ERROR_UNSUPPORTED_FEATURE;
diff --git a/source/adapters/mock/ur_mockddi.cpp b/source/adapters/mock/ur_mockddi.cpp
index 79e2b34b5f..baacaacb19 100644
--- a/source/adapters/mock/ur_mockddi.cpp
+++ b/source/adapters/mock/ur_mockddi.cpp
@@ -10003,9 +10003,13 @@ __urdlllocal ur_result_t UR_APICALL urEnqueueCooperativeKernelLaunchExp(
 /// @brief Intercept function for urKernelSuggestMaxCooperativeGroupCountExp
 __urdlllocal ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
     ur_kernel_handle_t hKernel, ///< [in] handle of the kernel object
-    size_t
-        localWorkSize, ///< [in] number of local work-items that will form a work-group when the
-                       ///< kernel is launched
+    uint32_t
+        workDim, ///< [in] number of dimensions, from 1 to 3, to specify the work-group
+                 ///< work-items
+    const size_t *
+        pLocalWorkSize, ///< [in] pointer to an array of workDim unsigned values that specify the
+    ///< number of local work-items forming a work-group that will execute the
+    ///< kernel function.
     size_t
         dynamicSharedMemorySize, ///< [in] size of dynamic shared memory, for each work-group, in bytes,
     ///< that will be used when the kernel is launched
@@ -10014,7 +10018,8 @@ __urdlllocal ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
     ur_result_t result = UR_RESULT_SUCCESS;
 
     ur_kernel_suggest_max_cooperative_group_count_exp_params_t params = {
-        &hKernel, &localWorkSize, &dynamicSharedMemorySize, &pGroupCountRet};
+        &hKernel, &workDim, &pLocalWorkSize, &dynamicSharedMemorySize,
+        &pGroupCountRet};
 
     auto beforeCallback = reinterpret_cast<ur_mock_callback_t>(
         mock::getCallbacks().get_before_callback(
diff --git a/source/adapters/opencl/kernel.cpp b/source/adapters/opencl/kernel.cpp
index f60c8a2715..df160b65eb 100644
--- a/source/adapters/opencl/kernel.cpp
+++ b/source/adapters/opencl/kernel.cpp
@@ -390,7 +390,8 @@ UR_APIEXPORT ur_result_t UR_APICALL urKernelGetNativeHandle(
 
 UR_APIEXPORT ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
     [[maybe_unused]] ur_kernel_handle_t hKernel,
-    [[maybe_unused]] size_t localWorkSize,
+    [[maybe_unused]] uint32_t workDim,
+    [[maybe_unused]] const size_t *pLocalWorkSize,
     [[maybe_unused]] size_t dynamicSharedMemorySize,
     [[maybe_unused]] uint32_t *pGroupCountRet) {
   return UR_RESULT_ERROR_UNSUPPORTED_FEATURE;
diff --git a/source/loader/layers/tracing/ur_trcddi.cpp b/source/loader/layers/tracing/ur_trcddi.cpp
index 7a110099da..bd620a159f 100644
--- a/source/loader/layers/tracing/ur_trcddi.cpp
+++ b/source/loader/layers/tracing/ur_trcddi.cpp
@@ -8585,9 +8585,13 @@ __urdlllocal ur_result_t UR_APICALL urEnqueueCooperativeKernelLaunchExp(
 /// @brief Intercept function for urKernelSuggestMaxCooperativeGroupCountExp
 __urdlllocal ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
     ur_kernel_handle_t hKernel, ///< [in] handle of the kernel object
-    size_t
-        localWorkSize, ///< [in] number of local work-items that will form a work-group when the
-                       ///< kernel is launched
+    uint32_t
+        workDim, ///< [in] number of dimensions, from 1 to 3, to specify the work-group
+                 ///< work-items
+    const size_t *
+        pLocalWorkSize, ///< [in] pointer to an array of workDim unsigned values that specify the
+    ///< number of local work-items forming a work-group that will execute the
+    ///< kernel function.
     size_t
         dynamicSharedMemorySize, ///< [in] size of dynamic shared memory, for each work-group, in bytes,
     ///< that will be used when the kernel is launched
@@ -8602,7 +8606,8 @@ __urdlllocal ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
     }
 
     ur_kernel_suggest_max_cooperative_group_count_exp_params_t params = {
-        &hKernel, &localWorkSize, &dynamicSharedMemorySize, &pGroupCountRet};
+        &hKernel, &workDim, &pLocalWorkSize, &dynamicSharedMemorySize,
+        &pGroupCountRet};
     uint64_t instance = getContext()->notify_begin(
         UR_FUNCTION_KERNEL_SUGGEST_MAX_COOPERATIVE_GROUP_COUNT_EXP,
         "urKernelSuggestMaxCooperativeGroupCountExp", &params);
@@ -8611,7 +8616,8 @@ __urdlllocal ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
     logger.info("   ---> urKernelSuggestMaxCooperativeGroupCountExp\n");
 
     ur_result_t result = pfnSuggestMaxCooperativeGroupCountExp(
-        hKernel, localWorkSize, dynamicSharedMemorySize, pGroupCountRet);
+        hKernel, workDim, pLocalWorkSize, dynamicSharedMemorySize,
+        pGroupCountRet);
 
     getContext()->notify_end(
         UR_FUNCTION_KERNEL_SUGGEST_MAX_COOPERATIVE_GROUP_COUNT_EXP,
diff --git a/source/loader/layers/validation/ur_valddi.cpp b/source/loader/layers/validation/ur_valddi.cpp
index 59bdd94de6..e27c5ae3e1 100644
--- a/source/loader/layers/validation/ur_valddi.cpp
+++ b/source/loader/layers/validation/ur_valddi.cpp
@@ -9613,9 +9613,13 @@ __urdlllocal ur_result_t UR_APICALL urEnqueueCooperativeKernelLaunchExp(
 /// @brief Intercept function for urKernelSuggestMaxCooperativeGroupCountExp
 __urdlllocal ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
     ur_kernel_handle_t hKernel, ///< [in] handle of the kernel object
-    size_t
-        localWorkSize, ///< [in] number of local work-items that will form a work-group when the
-                       ///< kernel is launched
+    uint32_t
+        workDim, ///< [in] number of dimensions, from 1 to 3, to specify the work-group
+                 ///< work-items
+    const size_t *
+        pLocalWorkSize, ///< [in] pointer to an array of workDim unsigned values that specify the
+    ///< number of local work-items forming a work-group that will execute the
+    ///< kernel function.
     size_t
         dynamicSharedMemorySize, ///< [in] size of dynamic shared memory, for each work-group, in bytes,
     ///< that will be used when the kernel is launched
@@ -9634,6 +9638,10 @@ __urdlllocal ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
             return UR_RESULT_ERROR_INVALID_NULL_HANDLE;
         }
 
+        if (NULL == pLocalWorkSize) {
+            return UR_RESULT_ERROR_INVALID_NULL_POINTER;
+        }
+
         if (NULL == pGroupCountRet) {
             return UR_RESULT_ERROR_INVALID_NULL_POINTER;
         }
@@ -9645,7 +9653,8 @@ __urdlllocal ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
     }
 
     ur_result_t result = pfnSuggestMaxCooperativeGroupCountExp(
-        hKernel, localWorkSize, dynamicSharedMemorySize, pGroupCountRet);
+        hKernel, workDim, pLocalWorkSize, dynamicSharedMemorySize,
+        pGroupCountRet);
 
     return result;
 }
diff --git a/source/loader/ur_ldrddi.cpp b/source/loader/ur_ldrddi.cpp
index 8134313a50..c482fbdfcc 100644
--- a/source/loader/ur_ldrddi.cpp
+++ b/source/loader/ur_ldrddi.cpp
@@ -8760,9 +8760,13 @@ __urdlllocal ur_result_t UR_APICALL urEnqueueCooperativeKernelLaunchExp(
 /// @brief Intercept function for urKernelSuggestMaxCooperativeGroupCountExp
 __urdlllocal ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
     ur_kernel_handle_t hKernel, ///< [in] handle of the kernel object
-    size_t
-        localWorkSize, ///< [in] number of local work-items that will form a work-group when the
-                       ///< kernel is launched
+    uint32_t
+        workDim, ///< [in] number of dimensions, from 1 to 3, to specify the work-group
+                 ///< work-items
+    const size_t *
+        pLocalWorkSize, ///< [in] pointer to an array of workDim unsigned values that specify the
+    ///< number of local work-items forming a work-group that will execute the
+    ///< kernel function.
     size_t
         dynamicSharedMemorySize, ///< [in] size of dynamic shared memory, for each work-group, in bytes,
     ///< that will be used when the kernel is launched
@@ -8785,7 +8789,8 @@ __urdlllocal ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
 
     // forward to device-platform
     result = pfnSuggestMaxCooperativeGroupCountExp(
-        hKernel, localWorkSize, dynamicSharedMemorySize, pGroupCountRet);
+        hKernel, workDim, pLocalWorkSize, dynamicSharedMemorySize,
+        pGroupCountRet);
 
     return result;
 }
diff --git a/source/loader/ur_libapi.cpp b/source/loader/ur_libapi.cpp
index 7a1840c2b9..46d3ca8eac 100644
--- a/source/loader/ur_libapi.cpp
+++ b/source/loader/ur_libapi.cpp
@@ -8893,13 +8893,18 @@ ur_result_t UR_APICALL urEnqueueCooperativeKernelLaunchExp(
 ///     - ::UR_RESULT_ERROR_INVALID_NULL_HANDLE
 ///         + `NULL == hKernel`
 ///     - ::UR_RESULT_ERROR_INVALID_NULL_POINTER
+///         + `NULL == pLocalWorkSize`
 ///         + `NULL == pGroupCountRet`
 ///     - ::UR_RESULT_ERROR_INVALID_KERNEL
 ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
     ur_kernel_handle_t hKernel, ///< [in] handle of the kernel object
-    size_t
-        localWorkSize, ///< [in] number of local work-items that will form a work-group when the
-                       ///< kernel is launched
+    uint32_t
+        workDim, ///< [in] number of dimensions, from 1 to 3, to specify the work-group
+                 ///< work-items
+    const size_t *
+        pLocalWorkSize, ///< [in] pointer to an array of workDim unsigned values that specify the
+    ///< number of local work-items forming a work-group that will execute the
+    ///< kernel function.
     size_t
         dynamicSharedMemorySize, ///< [in] size of dynamic shared memory, for each work-group, in bytes,
     ///< that will be used when the kernel is launched
@@ -8913,7 +8918,8 @@ ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
     }
 
     return pfnSuggestMaxCooperativeGroupCountExp(
-        hKernel, localWorkSize, dynamicSharedMemorySize, pGroupCountRet);
+        hKernel, workDim, pLocalWorkSize, dynamicSharedMemorySize,
+        pGroupCountRet);
 } catch (...) {
     return exceptionToResult(std::current_exception());
 }
diff --git a/source/ur_api.cpp b/source/ur_api.cpp
index f342e4efa2..e18cc776a5 100644
--- a/source/ur_api.cpp
+++ b/source/ur_api.cpp
@@ -7543,13 +7543,18 @@ ur_result_t UR_APICALL urEnqueueCooperativeKernelLaunchExp(
 ///     - ::UR_RESULT_ERROR_INVALID_NULL_HANDLE
 ///         + `NULL == hKernel`
 ///     - ::UR_RESULT_ERROR_INVALID_NULL_POINTER
+///         + `NULL == pLocalWorkSize`
 ///         + `NULL == pGroupCountRet`
 ///     - ::UR_RESULT_ERROR_INVALID_KERNEL
 ur_result_t UR_APICALL urKernelSuggestMaxCooperativeGroupCountExp(
     ur_kernel_handle_t hKernel, ///< [in] handle of the kernel object
-    size_t
-        localWorkSize, ///< [in] number of local work-items that will form a work-group when the
-                       ///< kernel is launched
+    uint32_t
+        workDim, ///< [in] number of dimensions, from 1 to 3, to specify the work-group
+                 ///< work-items
+    const size_t *
+        pLocalWorkSize, ///< [in] pointer to an array of workDim unsigned values that specify the
+    ///< number of local work-items forming a work-group that will execute the
+    ///< kernel function.
     size_t
         dynamicSharedMemorySize, ///< [in] size of dynamic shared memory, for each work-group, in bytes,
     ///< that will be used when the kernel is launched