diff --git a/Ironwood/src/benchmark_gemm.py b/Ironwood/src/benchmark_gemm.py
index c8c27bbe..0534b347 100644
--- a/Ironwood/src/benchmark_gemm.py
+++ b/Ironwood/src/benchmark_gemm.py
@@ -155,6 +155,7 @@ def gemm_multiple_run_calculate_metrics(
         total_flops,
         total_flops_all_devices,
         peak_flops,
+        dtype=dtype.dtype.name,
     )
 
 def gemm_simple(
diff --git a/Ironwood/src/benchmark_inference_compute.py b/Ironwood/src/benchmark_inference_compute.py
index 5059de12..8bfa6b0e 100644
--- a/Ironwood/src/benchmark_inference_compute.py
+++ b/Ironwood/src/benchmark_inference_compute.py
@@ -124,7 +124,7 @@ def add_calculate_metrics(
         total_bytes, SHARDING_STRATEGY
     )
     return unified_bytes_metrics(
-        m, n, time_ms_list, total_bytes, total_bytes_all_devices
+        m, n, time_ms_list, total_bytes, total_bytes_all_devices, dtype=dtype.dtype.name
     )
 
 
@@ -191,7 +191,7 @@ def rmsnorm_calculate_metrics(
         total_bytes, SHARDING_STRATEGY
     )
     return unified_bytes_metrics(
-        m, n, time_ms_list, total_bytes, total_bytes_all_devices
+        m, n, time_ms_list, total_bytes, total_bytes_all_devices, dtype=dtype.dtype.name
     )
 
 
@@ -264,7 +264,7 @@ def silu_mul_calculate_metrics(
         total_bytes, SHARDING_STRATEGY
     )
     return unified_bytes_metrics(
-        m, n, time_ms_list, total_bytes, total_bytes_all_devices
+        m, n, time_ms_list, total_bytes, total_bytes_all_devices, dtype=dtype.dtype.name
     )
 
 
@@ -325,7 +325,7 @@ def sigmoid_calculate_metrics(
         total_bytes, SHARDING_STRATEGY
     )
     return unified_bytes_metrics(
-        m, n, time_ms_list, total_bytes, total_bytes_all_devices
+        m, n, time_ms_list, total_bytes, total_bytes_all_devices, dtype=dtype.dtype.name
     )
 
 
@@ -413,4 +413,4 @@ def sigmoid_calculate_metrics(
 #     scale = 2 if dtype == jnp.bfloat16 else 1
 #     total_bytes = scale * 3 * m
 #     total_bytes, total_bytes_all_devices = handle_based_on_sharding(total_bytes, SHARDING_STRATEGY)
-#     return unified_bytes_metrics(m, 0,  time_ms_list, total_bytes, total_bytes_all_devices)
+#     return unified_bytes_metrics(m, 0,  time_ms_list, total_bytes, total_bytes_all_devices, dtype=dtype.dtype.name)
diff --git a/Ironwood/src/benchmark_utils.py b/Ironwood/src/benchmark_utils.py
index d0e72b9e..baa33b5f 100644
--- a/Ironwood/src/benchmark_utils.py
+++ b/Ironwood/src/benchmark_utils.py
@@ -1110,6 +1110,7 @@ def unified_flops_metrics(
     total_flops: int,
     total_flops_all_devices: int,
     peak_TFLOPS_per_device: float,
+    dtype: str = None,
 ) -> Dict[str, Any]:
     """Calculates the metrics for the naive matmul benchmark."""
     # Build dictionary of all the parameters in the function
@@ -1178,6 +1179,7 @@ def unified_bytes_metrics(
     total_bytes: int,
     total_bytes_all_devices: int = 1e9,
     quant_dtype: str = None,
+    dtype: str = None,
 ) -> Dict[str, Any]:
     """Calculates the metrics for the naive matmul benchmark."""
     # Build dictionary of all the parameters in the function
@@ -1212,6 +1214,9 @@ def unified_bytes_metrics(
     if quant_dtype is not None:
         metadata.update({"quant_dtype": quant_dtype})
         metrics.update({"quant_dtype": quant_dtype})
+    if dtype is not None:
+        metadata.update({"dtype": dtype})
+        metrics.update({"dtype": dtype})
     metadata.update(
         {
             "StepTime(median,ms)": average_time_ms_statistics.statistics["p50"],