diff --git a/cpp/kernels/contextAttentionKernels/cuteDslFMHARunner.cpp b/cpp/kernels/contextAttentionKernels/cuteDslFMHARunner.cpp
index f849741c..83bfee41 100644
--- a/cpp/kernels/contextAttentionKernels/cuteDslFMHARunner.cpp
+++ b/cpp/kernels/contextAttentionKernels/cuteDslFMHARunner.cpp
@@ -38,7 +38,7 @@ namespace
 
 bool isSupportedBlackwellFmha(int32_t smVersion)
 {
-    return smVersion == 100 || smVersion == 101 || smVersion == 110;
+    return smVersion == 100 || smVersion == 101 || smVersion == 103 || smVersion == 110;
 }
 
 template <auto moduleLoader, auto moduleUnloader, typename Module>
diff --git a/cpp/kernels/contextAttentionKernels/cuteDslFMHAV2Runner.cpp b/cpp/kernels/contextAttentionKernels/cuteDslFMHAV2Runner.cpp
index 40cb3700..39a79ffd 100644
--- a/cpp/kernels/contextAttentionKernels/cuteDslFMHAV2Runner.cpp
+++ b/cpp/kernels/contextAttentionKernels/cuteDslFMHAV2Runner.cpp
@@ -64,7 +64,8 @@ namespace
 bool isFMHAV2SM(int32_t smVersion)
 {
     return smVersion == 80 || smVersion == 86 || smVersion == 87 || smVersion == 89 || smVersion == 90
-        || smVersion == 100 || smVersion == 101 || smVersion == 110 || smVersion == 120 || smVersion == 121;
+        || smVersion == 100 || smVersion == 101 || smVersion == 103 || smVersion == 110 || smVersion == 120
+        || smVersion == 121;
 }
 
 bool isFMHAV2PagedLlmHeadSize(int32_t headSize)
diff --git a/cpp/kernels/decodeAttentionKernels/decoderXQAJitCompiler.cpp b/cpp/kernels/decodeAttentionKernels/decoderXQAJitCompiler.cpp
index 9922199c..c33a8013 100644
--- a/cpp/kernels/decodeAttentionKernels/decoderXQAJitCompiler.cpp
+++ b/cpp/kernels/decodeAttentionKernels/decoderXQAJitCompiler.cpp
@@ -251,7 +251,7 @@ bool canCompileXQAKernel(int32_t numQHeads, int32_t numKVHeads, int32_t headSize
     nvinfer1::DataType dataType, nvinfer1::DataType kvDataType) noexcept
 {
     // SM versions with an XQA decode kernel at all.
-    constexpr std::array<int32_t, 9> kALLOWED_SM_VERSIONS{80, 86, 87, 89, 90, 100, 101, 120, 121};
+    constexpr std::array<int32_t, 10> kALLOWED_SM_VERSIONS{80, 86, 87, 89, 90, 100, 101, 103, 120, 121};
     // Of those, the ones with native FP8 converts. Listed explicitly rather
     // than tested as `smVersion >= 89`: a numeric threshold silently admits
     // every future SM added to kALLOWED_SM_VERSIONS, and whether that SM has
diff --git a/kernelSrcs/build_cutedsl.py b/kernelSrcs/build_cutedsl.py
index b8fd3535..953bc89d 100644
--- a/kernelSrcs/build_cutedsl.py
+++ b/kernelSrcs/build_cutedsl.py
@@ -264,14 +264,14 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d64",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,64", "--k_shape", "1,1024,1,64"] + _LLM,
     ),
     KernelVariant(
         name="fmha_d128",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,128", "--k_shape", "1,1024,1,128"] + _LLM,
     ),
@@ -282,7 +282,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d64_skipsoftmax",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,64", "--k_shape", "1,1024,1,64"]
                     + _LLM + ["--skip_softmax_threshold", "1e-6"],
@@ -290,7 +290,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d128_skipsoftmax",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,128", "--k_shape", "1,1024,1,128"]
                     + _LLM + ["--skip_softmax_threshold", "1e-6"],
@@ -298,7 +298,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d256",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,16,256", "--k_shape", "1,1024,2,256"] + _LLM,
     ),
@@ -307,14 +307,14 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d512_paged",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,8,512", "--k_shape", "1,1024,1,512"] + _LLM_PAGED,
     ),
     KernelVariant(
         name="fmha_d512_paged_bidirectional",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,8,512", "--k_shape", "1,1024,1,512"]
                     + _LLM_PAGED + ["--window_size", "4096,-1", "--bidirectional"],
@@ -322,7 +322,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d512_sw_paged",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,8,512", "--k_shape", "1,1024,1,512"]
                     + _LLM_PAGED + ["--window_size", "4096,-1"],
@@ -330,28 +330,28 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d512_paged_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,8,512", "--k_shape", "1,1024,1,512"] + _LLM_FP8_PAGED,
     ),
     KernelVariant(
         name="fmha_d512_dense_paged",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,8,512", "--k_shape", "1,1024,1,512"] + _LLM_DENSE_PAGED,
     ),
     KernelVariant(
         name="fmha_d512_dense_paged_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,8,512", "--k_shape", "1,1024,1,512"] + _LLM_DENSE_FP8_PAGED,
     ),
     KernelVariant(
         name="fmha_d512_sw_paged_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,8,512", "--k_shape", "1,1024,1,512"]
                     + _LLM_FP8_PAGED + ["--window_size", "4096,-1"],
@@ -359,7 +359,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d64_sw",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,64", "--k_shape", "1,1024,1,64"]
                     + _LLM + ["--window_size", "4096,-1"],
@@ -367,7 +367,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d128_sw",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,128", "--k_shape", "1,1024,1,128"]
                     + _LLM + ["--window_size", "4096,-1"],
@@ -375,7 +375,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d256_sw",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,16,256", "--k_shape", "1,1024,2,256"]
                     + _LLM + ["--window_size", "4096,-1"],
@@ -384,28 +384,28 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d64_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,64", "--k_shape", "1,1024,1,64"] + _LLM_FP8,
     ),
     KernelVariant(
         name="fmha_d128_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,128", "--k_shape", "1,1024,1,128"] + _LLM_FP8,
     ),
     KernelVariant(
         name="fmha_d256_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,16,256", "--k_shape", "1,1024,2,256"] + _LLM_FP8,
     ),
     KernelVariant(
         name="fmha_d64_sw_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,64", "--k_shape", "1,1024,1,64"]
                     + _LLM_FP8 + ["--window_size", "4096,-1"],
@@ -413,7 +413,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d128_sw_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,128", "--k_shape", "1,1024,1,128"]
                     + _LLM_FP8 + ["--window_size", "4096,-1"],
@@ -421,7 +421,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d256_sw_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,16,256", "--k_shape", "1,1024,2,256"]
                     + _LLM_FP8 + ["--window_size", "4096,-1"],
@@ -430,14 +430,14 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d64_paged",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,64", "--k_shape", "1,1024,1,64"] + _LLM_PAGED,
     ),
     KernelVariant(
         name="fmha_d128_paged",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,128", "--k_shape", "1,1024,1,128"] + _LLM_PAGED,
     ),
@@ -447,7 +447,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d64_skipsoftmax_paged",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,64", "--k_shape", "1,1024,1,64"]
                     + _LLM_PAGED + ["--skip_softmax_threshold", "1e-6"],
@@ -455,7 +455,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d128_skipsoftmax_paged",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,128", "--k_shape", "1,1024,1,128"]
                     + _LLM_PAGED + ["--skip_softmax_threshold", "1e-6"],
@@ -463,21 +463,21 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d256_paged",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,16,256", "--k_shape", "1,1024,2,256"] + _LLM_PAGED,
     ),
     KernelVariant(
         name="fmha_d256_dense_paged",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,16,256", "--k_shape", "1,1024,2,256"] + _LLM_DENSE_PAGED,
     ),
     KernelVariant(
         name="fmha_d64_sw_paged",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,64", "--k_shape", "1,1024,1,64"]
                     + _LLM_PAGED + ["--window_size", "4096,-1"],
@@ -485,7 +485,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d128_sw_paged",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,128", "--k_shape", "1,1024,1,128"]
                     + _LLM_PAGED + ["--window_size", "4096,-1"],
@@ -494,7 +494,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d256_sw_paged",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,16,256", "--k_shape", "1,1024,2,256"]
                     + _LLM_PAGED + ["--window_size", "4096,-1"],
@@ -502,35 +502,35 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d64_paged_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,64", "--k_shape", "1,1024,1,64"] + _LLM_FP8_PAGED,
     ),
     KernelVariant(
         name="fmha_d128_paged_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,128", "--k_shape", "1,1024,1,128"] + _LLM_FP8_PAGED,
     ),
     KernelVariant(
         name="fmha_d256_paged_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,16,256", "--k_shape", "1,1024,2,256"] + _LLM_FP8_PAGED,
     ),
     KernelVariant(
         name="fmha_d256_dense_paged_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,16,256", "--k_shape", "1,1024,2,256"] + _LLM_DENSE_FP8_PAGED,
     ),
     KernelVariant(
         name="fmha_d64_sw_paged_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,64", "--k_shape", "1,1024,1,64"]
                     + _LLM_FP8_PAGED + ["--window_size", "4096,-1"],
@@ -538,7 +538,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d128_sw_paged_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,128", "--k_shape", "1,1024,1,128"]
                     + _LLM_FP8_PAGED + ["--window_size", "4096,-1"],
@@ -546,7 +546,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_d256_sw_paged_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,16,256", "--k_shape", "1,1024,2,256"]
                     + _LLM_FP8_PAGED + ["--window_size", "4096,-1"],
@@ -554,35 +554,35 @@ class KernelVariant:
     KernelVariant(
         name="vit_fmha_d64",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,64", "--k_shape", "1,1024,14,64"] + _VIT,
     ),
     KernelVariant(
         name="vit_fmha_d72",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,72", "--k_shape", "1,1024,14,72"] + _VIT,
     ),
     KernelVariant(
         name="vit_fmha_d80",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,80", "--k_shape", "1,1024,14,80"] + _VIT,
     ),
     KernelVariant(
         name="vit_fmha_d96",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,96", "--k_shape", "1,1024,14,96"] + _VIT,
     ),
     KernelVariant(
         name="vit_fmha_d128",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,128", "--k_shape", "1,1024,14,128"] + _VIT,
     ),
@@ -594,14 +594,14 @@ class KernelVariant:
     KernelVariant(
         name="vit_fmha_d64_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,64", "--k_shape", "1,1024,14,64"] + _VIT_FP8 + ["--kv_stage", "2"],
     ),
     KernelVariant(
         name="vit_fmha_d80_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         # Pinned to the measured optimum: the padded K=96 smem tiles need a
         # deeper KV pipeline than the narrower dims (shallow staging costs
@@ -611,14 +611,14 @@ class KernelVariant:
     KernelVariant(
         name="vit_fmha_d96_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,96", "--k_shape", "1,1024,14,96"] + _VIT_FP8,
     ),
     KernelVariant(
         name="vit_fmha_d128_fp8",
         group="fmha",
-        supported_sms=[100, 101, 110],
+        supported_sms=[100, 101, 103, 110],
         script="fmha_cutedsl_blackwell/fmha.py",
         script_args=["--q_shape", "1,1024,14,128", "--k_shape", "1,1024,14,128"] + _VIT_FP8 + ["--kv_stage", "3"],
     ),
@@ -631,7 +631,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d64",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "64",
@@ -644,7 +644,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d64_small",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "64",
@@ -655,7 +655,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d128",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "128",
@@ -666,7 +666,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d256",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "256",
@@ -677,7 +677,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d64_sw",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "64",
@@ -689,7 +689,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d128_sw",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "128",
@@ -701,7 +701,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d256_sw",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "256",
@@ -713,7 +713,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d64_paged",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "64",
@@ -724,7 +724,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d64_small_paged",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "64",
@@ -735,7 +735,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d128_paged",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "128",
@@ -746,7 +746,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d256_paged",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "256",
@@ -757,7 +757,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d512_paged",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "512",
@@ -768,7 +768,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d64_sw_paged",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "64",
@@ -780,7 +780,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d128_sw_paged",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "128",
@@ -792,7 +792,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d256_sw_paged",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "256",
@@ -804,7 +804,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d512_sw_paged",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "512",
@@ -816,7 +816,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d256_padding",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "256",
@@ -827,7 +827,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_vit_d64",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "64",
@@ -838,7 +838,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_vit_d72",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "72",
@@ -849,7 +849,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_vit_d80",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "80",
@@ -860,7 +860,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_vit_d128",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "128",
@@ -871,7 +871,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d256_bidirectional",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "256",
@@ -889,7 +889,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d512",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "512",
@@ -901,7 +901,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d512_sw",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "512",
@@ -919,7 +919,7 @@ class KernelVariant:
     KernelVariant(
         name="fmha_v2_d512_bidirectional",
         group="fmha",
-        supported_sms=[80, 86, 87, 89, 90, 100, 101, 110, 120, 121],
+        supported_sms=[80, 86, 87, 89, 90, 100, 101, 103, 110, 120, 121],
         script="fmha_v2_cutedsl/fmha.py",
         script_args=[
             "--head_dim", "512",
@@ -1755,7 +1755,7 @@ class KernelVariant:
         KernelVariant(
             name=f"int4_fp16_gemm_{_tile_tag}_s{_stages}_sk{_sk}",
             group="int4_fp16_gemm",
-            supported_sms=[80, 86, 87, 89, 100, 101, 110, 120, 121],
+            supported_sms=[80, 86, 87, 89, 100, 101, 103, 110, 120, 121],
             script="int4_fp16_gemm_cutedsl/int4_fp16_gemm_ampere.py",
             script_args=[
                 "--mnk", "256,512,1024",
@@ -1788,7 +1788,7 @@ class KernelVariant:
         KernelVariant(
             name=f"int4_fp16_gemv_m{_m}",
             group="int4_fp16_gemm",
-            supported_sms=[80, 86, 87, 89, 100, 101, 110, 120, 121],
+            supported_sms=[80, 86, 87, 89, 100, 101, 103, 110, 120, 121],
             script="int4_fp16_gemm_cutedsl/int4_fp16_gemv_ampere.py",
             script_args=[
                 "--mnk", f"{_m},512,1024",
diff --git a/kernelSrcs/xqa/mha.cu b/kernelSrcs/xqa/mha.cu
index 159f2fc3..7248547a 100644
--- a/kernelSrcs/xqa/mha.cu
+++ b/kernelSrcs/xqa/mha.cu
@@ -102,7 +102,7 @@ constexpr uint32_t preferedKHeadPartBytes = tiledQKVStagingHeadDim512 ? 32 : 64;
 __constant__ constexpr uint32_t cacheVTileSeqLen
     = tiledQKVStagingHeadDim512 ? tiledQKVStagingHeadDim512VTileSeqLen : 32;
 #elif __CUDA_ARCH__ == 800 || __CUDA_ARCH__ == 870 || __CUDA_ARCH__ == 900 || __CUDA_ARCH__ == 1000                    \
-    || __CUDA_ARCH__ == 1010 || __CUDA_ARCH__ == 1100
+    || __CUDA_ARCH__ == 1010 || __CUDA_ARCH__ == 1030 || __CUDA_ARCH__ == 1100
 // This path tiles Q/K/V shared-memory staging to keep shared memory within the per-CTA budget.
 constexpr uint32_t preferedKHeadPartBytes = tiledQKVStagingHeadDim512 ? 32 : (useSmallHeadDim512Tiles ? 64 : 128);
 constexpr bool useWide2CtaHeadDim512VTile = twoCtaHeadDim512 && cacheElemSize < inputElemSize;
diff --git a/kernelSrcs/xqa/utils.cuh b/kernelSrcs/xqa/utils.cuh
index 40669db1..aa2e7513 100644
--- a/kernelSrcs/xqa/utils.cuh
+++ b/kernelSrcs/xqa/utils.cuh
@@ -51,7 +51,8 @@ inline constexpr bool isLowPrecCacheElem = mha::is_same_v<T, int8_t> || isFp8E4M
 constexpr uint32_t kMAX_SMEM_SIZE = (99u << 10);
 #elif __CUDA_ARCH__ == 800 || __CUDA_ARCH__ == 870
 constexpr uint32_t kMAX_SMEM_SIZE = (163u << 10);
-#elif __CUDA_ARCH__ == 900 || __CUDA_ARCH__ == 1000 || __CUDA_ARCH__ == 1010 || __CUDA_ARCH__ == 1100
+#elif __CUDA_ARCH__ == 900 || __CUDA_ARCH__ == 1000 || __CUDA_ARCH__ == 1010 || __CUDA_ARCH__ == 1030          \
+    || __CUDA_ARCH__ == 1100
 constexpr uint32_t kMAX_SMEM_SIZE = (227u << 10);
 #endif
 #endif