diff --git a/webnn/conformance_tests/dequantizeLinear.https.any.js b/webnn/conformance_tests/dequantizeLinear.https.any.js
index 73dd916b976316..6b8ba371693644 100644
--- a/webnn/conformance_tests/dequantizeLinear.https.any.js
+++ b/webnn/conformance_tests/dequantizeLinear.https.any.js
@@ -119,8 +119,8 @@ const dequantizeLinearTests = [
           'constant': true
         },
         'dequantizeLinearZeroPoint': {
-          'data': [128],
-          'descriptor': {shape: [], dataType: 'uint8'},
+          'data': [128, 128, 128, 128],
+          'descriptor': {shape: [4], dataType: 'uint8'},
           'constant': true
         }
       },
@@ -144,6 +144,50 @@ const dequantizeLinearTests = [
       }
     }
   },
+  {
+    'name':
+        'dequantizeLinear uint8 1D constant tensor with implicit block_size = 2.',
+    'graph': {
+      'inputs': {
+        'dequantizeLinearInput': {
+          'data': [12, 24, 35, 123],
+          'descriptor': {shape: [4], dataType: 'uint8'},
+          'constant': true
+        },
+        'dequantizeLinearScale': {
+          'data': [
+            9.343092918395996,
+            -4.617084980010986,
+          ],
+          'descriptor': {shape: [2], dataType: 'float32'},
+          'constant': true
+        },
+        'dequantizeLinearZeroPoint': {
+          'data': [128, 110],
+          'descriptor': {shape: [2], dataType: 'uint8'},
+          'constant': true
+        }
+      },
+      'operators': [{
+        'name': 'dequantizeLinear',
+        'arguments': [
+          {'input': 'dequantizeLinearInput'},
+          {'scale': 'dequantizeLinearScale'},
+          {'zeroPoint': 'dequantizeLinearZeroPoint'}
+        ],
+        'outputs': 'dequantizeLinearOutput'
+      }],
+      'expectedOutputs': {
+        'dequantizeLinearOutput': {
+          'data': [
+            -1083.798828125, -971.681640625, 346.2813720703125,
+            -60.0221061706543
+          ],
+          'descriptor': {shape: [4], dataType: 'float32'}
+        }
+      }
+    }
+  },
   {
     'name':
         'dequantizeLinear int8 4D constant tensor broadcasting scale and zeroPoint',
@@ -160,8 +204,8 @@ const dequantizeLinearTests = [
           'constant': true
         },
         'dequantizeLinearZeroPoint': {
-          'data': [12],
-          'descriptor': {shape: [], dataType: 'int8'},
+          'data': [12, 12],
+          'descriptor': {shape: [2, 1], dataType: 'int8'},
           'constant': true
         }
       },
@@ -185,6 +229,74 @@ const dequantizeLinearTests = [
       }
     }
   },
+  {
+    'name': 'dequantizeLinear int8 4D constant tensor with block_size = [3, 2]',
+    'graph': {
+      'inputs': {
+        'dequantizeLinearInput': {
+          'data': [
+            -124, 0,   23,  122, 12, 23, 45, 36, 67, 78, -22, 0,
+            -34,  -45, -56, -67, 89, 30, 12, 23, 56, 67, 56,  -12
+          ],
+          'descriptor': {shape: [6, 4], dataType: 'int8'},
+          'constant': true
+        },
+        'dequantizeLinearScale': {
+          'data': [
+            0.2800687253475189, -4.617084980010986, 1.2800687253475189,
+            -3.617084980010986
+          ],
+          'descriptor': {shape: [2, 2], dataType: 'float32'},
+          'constant': true
+        },
+        'dequantizeLinearZeroPoint': {
+          'data': [1, 3, 5, 12],
+          'descriptor': {shape: [2, 2], dataType: 'int8'},
+          'constant': true
+        }
+      },
+      'operators': [{
+        'name': 'dequantizeLinear',
+        'arguments': [
+          {'input': 'dequantizeLinearInput'},
+          {'scale': 'dequantizeLinearScale'},
+          {'zeroPoint': 'dequantizeLinearZeroPoint'}
+        ],
+        'outputs': 'dequantizeLinearOutput'
+      }],
+      'expectedOutputs': {
+        'dequantizeLinearOutput': {
+          'data': [
+            -35.00859069824219,
+            -0.2800687253475189,
+            -92.3416976928711,
+            -549.43310546875,
+            3.0807559490203857,
+            6.1615118980407715,
+            -193.91757202148438,
+            -152.36380004882812,
+            18.484535217285156,
+            21.565292358398438,
+            115.4271240234375,
+            13.851255416870117,
+            -49.92267990112305,
+            -64.0034408569336,
+            245.96177673339844,
+            285.7497253417969,
+            107.52577209472656,
+            32.0017204284668,
+            0,
+            -39.787933349609375,
+            65.28350830078125,
+            79.36426544189453,
+            -159.1517333984375,
+            86.81004333496094
+          ],
+          'descriptor': {shape: [6, 4], dataType: 'float32'}
+        }
+      }
+    }
+  },
   {
     'name': 'dequantizeLinear uint4 1D tensor with even input size',
     'graph': {
@@ -200,8 +312,8 @@ const dequantizeLinearTests = [
           'constant': true
         },
         'dequantizeLinearZeroPoint': {
-          'data': [0],
-          'descriptor': {shape: [], dataType: 'uint4'},
+          'data': [0, 1],
+          'descriptor': {shape: [2], dataType: 'uint4'},
           'constant': true
         }
       },
@@ -216,7 +328,7 @@ const dequantizeLinearTests = [
       }],
       'expectedOutputs': {
         'dequantizeLinearOutput': {
-          'data': [16.804121017456055, 0],
+          'data': [16.804121017456055, -1.1202747821807861],
           'descriptor': {shape: [2], dataType: 'float32'}
         }
       }
@@ -237,8 +349,8 @@ const dequantizeLinearTests = [
           'constant': true
         },
         'dequantizeLinearZeroPoint': {
-          'data': [2, 1, 4],
-          'descriptor': {shape: [3], dataType: 'uint4'},
+          'data': [2],
+          'descriptor': {shape: [1], dataType: 'uint4'},
           'constant': true
         }
       },
@@ -253,7 +365,7 @@ const dequantizeLinearTests = [
       }],
       'expectedOutputs': {
         'dequantizeLinearOutput': {
-          'data': [8.962198257446289, 12.323022842407227, 11.202747344970703],
+          'data': [8.962198257446289, 11.202747344970703, 13.443297386169434],
           'descriptor': {shape: [3], dataType: 'float32'}
         }
       }
@@ -278,7 +390,7 @@ const dequantizeLinearTests = [
         },
         'dequantizeLinearZeroPoint': {
           'data': [2, 3],
-          'descriptor': {shape: [2], dataType: 'uint4'},
+          'descriptor': {shape: [2, 1], dataType: 'uint4'},
           'constant': true
         }
       },
@@ -294,7 +406,7 @@ const dequantizeLinearTests = [
       'expectedOutputs': {
         'dequantizeLinearOutput': {
           'data': [
-            -18.686185836791992, -18.686185836791992, -36.93667984008789,
+            -18.686185836791992, -9.343092918395996, -32.31959533691406,
             -55.40502166748047
           ],
           'descriptor': {shape: [1, 1, 2, 2], dataType: 'float32'}
@@ -302,6 +414,49 @@ const dequantizeLinearTests = [
       }
     }
   },
+  {
+    'name': 'dequantizeLinear uint4 3D input with block_size = [1, 1, 2]',
+    'graph': {
+      'inputs': {
+        'dequantizeLinearInput': {
+          'data': [0, 1, 10, 15],
+          'descriptor': {shape: [1, 1, 4], dataType: 'uint4'},
+          'constant': true
+        },
+        'dequantizeLinearScale': {
+          'data': [
+            9.343092918395996,
+            -4.617084980010986,
+          ],
+          'descriptor': {shape: [1, 2], dataType: 'float32'},
+          'constant': true
+        },
+        'dequantizeLinearZeroPoint': {
+          'data': [2, 3],
+          'descriptor': {shape: [1, 2], dataType: 'uint4'},
+          'constant': true
+        }
+      },
+      'operators': [{
+        'name': 'dequantizeLinear',
+        'arguments': [
+          {'input': 'dequantizeLinearInput'},
+          {'scale': 'dequantizeLinearScale'},
+          {'zeroPoint': 'dequantizeLinearZeroPoint'}
+        ],
+        'outputs': 'dequantizeLinearOutput'
+      }],
+      'expectedOutputs': {
+        'dequantizeLinearOutput': {
+          'data': [
+            -18.686185836791992, -9.343092918395996, -32.31959533691406,
+            -55.40502166748047
+          ],
+          'descriptor': {shape: [1, 1, 4], dataType: 'float32'}
+        }
+      }
+    }
+  },
   {
     'name': 'dequantizeLinear int4 1D tensor with even size',
     'graph': {
@@ -312,8 +467,8 @@ const dequantizeLinearTests = [
           'constant': true
         },
         'dequantizeLinearScale': {
-          'data': [1.1202747821807861],
-          'descriptor': {shape: [], dataType: 'float32'},
+          'data': [1.1202747821807861, 1.1202747821807861],
+          'descriptor': {shape: [2], dataType: 'float32'},
           'constant': true
         },
         'dequantizeLinearZeroPoint': {
@@ -350,12 +505,12 @@ const dequantizeLinearTests = [
         },
         'dequantizeLinearScale': {
           'data': [1.1202747821807861],
-          'descriptor': {shape: [], dataType: 'float32'},
+          'descriptor': {shape: [1], dataType: 'float32'},
           'constant': true
         },
         'dequantizeLinearZeroPoint': {
-          'data': [-3, 0, 0],
-          'descriptor': {shape: [3], dataType: 'int4'},
+          'data': [-3],
+          'descriptor': {shape: [1], dataType: 'int4'},
           'constant': true
         }
       },
@@ -370,7 +525,7 @@ const dequantizeLinearTests = [
       }],
       'expectedOutputs': {
         'dequantizeLinearOutput': {
-          'data': [2.2405495643615723, 7.841923713684082, 0],
+          'data': [2.2405495643615723, 11.202747344970703, 3.3608243465423584],
           'descriptor': {shape: [3], dataType: 'float32'}
         }
       }
diff --git a/webnn/conformance_tests/quantizeLinear.https.any.js b/webnn/conformance_tests/quantizeLinear.https.any.js
index 8aa9d7f3bcc407..8333f7b1b58a0a 100644
--- a/webnn/conformance_tests/quantizeLinear.https.any.js
+++ b/webnn/conformance_tests/quantizeLinear.https.any.js
@@ -83,8 +83,8 @@ const quantizeLinearTests = [
           'constant': true
         },
         'quantizeLinearZeroPoint': {
-          'data': [128],
-          'descriptor': {shape: [], dataType: 'uint8'},
+          'data': [128, 128, 128, 128],
+          'descriptor': {shape: [4], dataType: 'uint8'},
           'constant': true
         }
       },
@@ -104,6 +104,46 @@ const quantizeLinearTests = [
       }
     }
   },
+  {
+    'name':
+        'quantizeLinear float32 2D constant tensor broadcasting zeroPoint and scale',
+    'graph': {
+      'inputs': {
+        'quantizeLinearInput': {
+          'data': [
+            -2.549168109893799, -4.794857501983643, 8.413617134094238,
+            6.108623504638672
+          ],
+          'descriptor': {shape: [2, 2], dataType: 'float32'},
+          'constant': true
+        },
+        'quantizeLinearScale': {
+          'data': [9.343092918395996],
+          'descriptor': {shape: [1], dataType: 'float32'},
+          'constant': true
+        },
+        'quantizeLinearZeroPoint': {
+          'data': [128],
+          'descriptor': {shape: [1], dataType: 'uint8'},
+          'constant': true
+        }
+      },
+      'operators': [{
+        'name': 'quantizeLinear',
+        'arguments': [
+          {'input': 'quantizeLinearInput'}, {'scale': 'quantizeLinearScale'},
+          {'zeroPoint': 'quantizeLinearZeroPoint'}
+        ],
+        'outputs': 'quantizeLinearOutput'
+      }],
+      'expectedOutputs': {
+        'quantizeLinearOutput': {
+          'data': [128, 127, 129, 129],
+          'descriptor': {shape: [2, 2], dataType: 'uint8'}
+        }
+      }
+    }
+  },
   {
     'name':
         'quantizeLinear float32 4D constant tensor broadcasting scale and zeroPoint',
@@ -123,8 +163,8 @@ const quantizeLinearTests = [
           'constant': true
         },
         'quantizeLinearZeroPoint': {
-          'data': [128],
-          'descriptor': {shape: [], dataType: 'uint8'},
+          'data': [128, 128],
+          'descriptor': {shape: [2, 1], dataType: 'uint8'},
           'constant': true
         }
       },
@@ -144,6 +184,46 @@ const quantizeLinearTests = [
       }
     }
   },
+  {
+    'name':
+        'quantizeLinear float32 3D input with implicit block_size = [1, 2, 1].',
+    'graph': {
+      'inputs': {
+        'quantizeLinearInput': {
+          'data': [
+            -2.549168109893799, -4.794857501983643, 8.413617134094238,
+            6.108623504638672
+          ],
+          'descriptor': {shape: [1, 4, 1], dataType: 'float32'},
+          'constant': true
+        },
+        'quantizeLinearScale': {
+          'data': [0.2800687253475189, -4.617084980010986],
+          'descriptor': {shape: [2, 1], dataType: 'float32'},
+          'constant': true
+        },
+        'quantizeLinearZeroPoint': {
+          'data': [128, 189],
+          'descriptor': {shape: [2, 1], dataType: 'uint8'},
+          'constant': true
+        }
+      },
+      'operators': [{
+        'name': 'quantizeLinear',
+        'arguments': [
+          {'input': 'quantizeLinearInput'}, {'scale': 'quantizeLinearScale'},
+          {'zeroPoint': 'quantizeLinearZeroPoint'}
+        ],
+        'outputs': 'quantizeLinearOutput'
+      }],
+      'expectedOutputs': {
+        'quantizeLinearOutput': {
+          'data': [119, 111, 187, 188],
+          'descriptor': {shape: [1, 4, 1], dataType: 'uint8'}
+        }
+      }
+    }
+  },
   {
     'name':
         'quantizeLinear float32 tensor with int4 zeroPoint which has odd size',
@@ -190,8 +270,8 @@ const quantizeLinearTests = [
           'constant': true
         },
         'quantizeLinearScale': {
-          'data': [1.1202747821807861],
-          'descriptor': {shape: [], dataType: 'float32'},
+          'data': [1.1202747821807861, 1.1202747821807861],
+          'descriptor': {shape: [2], dataType: 'float32'},
           'constant': true
         },
         'quantizeLinearZeroPoint': {
@@ -254,6 +334,47 @@ const quantizeLinearTests = [
       }
     }
   },
+  {
+    'name': 'quantizeLinear int4 zeroPoint with block_size = [3, 2]',
+    'graph': {
+      'inputs': {
+        'quantizeLinearInput': {
+          'data': [
+            4.794857501983643, 3.23434354545, 2.794857501983643,
+            5.794857501983643, 0, 7.23434354545, 4.794857501983643,
+            3.23434354545, 2.794857501983643, 5.794857501983643, 0,
+            7.23434354545
+          ],
+          'descriptor': {shape: [3, 4], dataType: 'float32'},
+          'constant': true
+        },
+        'quantizeLinearScale': {
+          'data': [1.1202747821807861, 2.1202747821807861],
+          'descriptor': {shape: [2], dataType: 'float32'},
+          'constant': true
+        },
+        'quantizeLinearZeroPoint': {
+          'data': [-6, -5],
+          'descriptor': {shape: [2], dataType: 'int4'},
+          'constant': true
+        }
+      },
+      'operators': [{
+        'name': 'quantizeLinear',
+        'arguments': [
+          {'input': 'quantizeLinearInput'}, {'scale': 'quantizeLinearScale'},
+          {'zeroPoint': 'quantizeLinearZeroPoint'}
+        ],
+        'outputs': 'quantizeLinearOutput'
+      }],
+      'expectedOutputs': {
+        'quantizeLinearOutput': {
+          'data': [-2, -3, -4, -3, -5, 0, -2, -3, -4, -1, -5, -2],
+          'descriptor': {shape: [3, 4], dataType: 'int4'}
+        }
+      }
+    }
+  },
   {
     'name':
         'quantizeLinear float32 tensor with uint4 zeroPoint which has odd size',
@@ -305,8 +426,8 @@ const quantizeLinearTests = [
           'constant': true
         },
         'quantizeLinearScale': {
-          'data': [1.1202747821807861],
-          'descriptor': {shape: [], dataType: 'float32'},
+          'data': [1.1202747821807861, 1.1202747821807861],
+          'descriptor': {shape: [2], dataType: 'float32'},
           'constant': true
         },
         'quantizeLinearZeroPoint': {
@@ -328,6 +449,45 @@ const quantizeLinearTests = [
             {'data': [5, 8], 'descriptor': {shape: [2], dataType: 'uint4'}}
       }
     }
+  },
+  {
+    'name': 'quantizeLinear uint4 zeroPoint with block_size = 3',
+    'graph': {
+      'inputs': {
+        'quantizeLinearInput': {
+          'data': [
+            4.794857501983643, 3.23434354545, 1.794857501983643, 2.23434354545,
+            4.794857501983643, 3.23434354545
+          ],
+          'descriptor': {shape: [6], dataType: 'float32'},
+          'constant': true
+        },
+        'quantizeLinearScale': {
+          'data': [1.1202747821807861, 1.1202747821807861],
+          'descriptor': {shape: [2], dataType: 'float32'},
+          'constant': true
+        },
+        'quantizeLinearZeroPoint': {
+          'data': [1, 5],
+          'descriptor': {shape: [2], dataType: 'uint4'},
+          'constant': true
+        }
+      },
+      'operators': [{
+        'name': 'quantizeLinear',
+        'arguments': [
+          {'input': 'quantizeLinearInput'}, {'scale': 'quantizeLinearScale'},
+          {'zeroPoint': 'quantizeLinearZeroPoint'}
+        ],
+        'outputs': 'quantizeLinearOutput'
+      }],
+      'expectedOutputs': {
+        'quantizeLinearOutput': {
+          'data': [5, 4, 3, 7, 9, 8],
+          'descriptor': {shape: [6], dataType: 'uint4'}
+        }
+      }
+    }
   }
 ];
 
diff --git a/webnn/validation_tests/dequantizeLinear.https.any.js b/webnn/validation_tests/dequantizeLinear.https.any.js
index d0dd2c15dfff2d..672fc4ac72456b 100644
--- a/webnn/validation_tests/dequantizeLinear.https.any.js
+++ b/webnn/validation_tests/dequantizeLinear.https.any.js
@@ -34,14 +34,22 @@ const tests = [
   },
   {
     name:
-        '[dequantizeLinear] Throw if the shape of scale is not broadcastable to the shape of input.',
+        '[dequantizeLinear] Test block-wise quantization with block_size = [2, 2, 5].',
+    input: {dataType: 'uint8', shape: [6, 4, 5]},
+    scale: {dataType: 'float32', shape: [3, 2, 1]},
+    zeroPoint: {dataType: 'uint8', shape: [3, 2, 1]},
+    output: {dataType: 'float32', shape: [6, 4, 5]},
+  },
+  {
+    name:
+        '[dequantizeLinear] Throw if the scale size is not a factor of input size.',
     input: {dataType: 'uint8', shape: [3, 2, 5]},
     scale: {dataType: 'float32', shape: [2]},
-    zeroPoint: {dataType: 'uint8', shape: [5]},
+    zeroPoint: {dataType: 'uint8', shape: [2]},
   },
   {
     name:
-        '[dequantizeLinear] Throw if the shape of zero_point is not broadcastable to the shape of input.',
+        '[dequantizeLinear] Throw if the shape of zero_point is not the same as the shape of input.',
     input: {dataType: 'uint8', shape: [3, 2, 5]},
     scale: {dataType: 'float32', shape: [5]},
     zeroPoint: {dataType: 'uint8', shape: [2]},
@@ -55,14 +63,14 @@ const tests = [
   },
   {
     name:
-        '[dequantizeLinear] Throw if the data type of input is not int8 or uint8.',
+        '[dequantizeLinear] Throw if the data type of input is not one of {int4, uint4, int8, uint8}.',
     input: {dataType: 'float16', shape: [3, 2, 5]},
     scale: {dataType: 'float32', shape: [5]},
     zeroPoint: {dataType: 'int8', shape: [5]},
   },
   {
     name:
-        '[dequantizeLinear] Throw if the data type of zero_point is not int8 or uint8.',
+        '[dequantizeLinear] Throw if the data type of zero_point is not one of {int4, uint4, int8, uint8}.',
     input: {dataType: 'int8', shape: [3, 2, 5]},
     scale: {dataType: 'float32', shape: [5]},
     zeroPoint: {dataType: 'int32', shape: [5]},
diff --git a/webnn/validation_tests/quantizeLinear.https.any.js b/webnn/validation_tests/quantizeLinear.https.any.js
index 42b8ac4b1f2c36..8516e9cf8e0b47 100644
--- a/webnn/validation_tests/quantizeLinear.https.any.js
+++ b/webnn/validation_tests/quantizeLinear.https.any.js
@@ -34,33 +34,43 @@ const tests = [
   },
   {
     name:
-        '[quantizeLinear] Throw if the shape of scale is not broadcastable to the shape of input.',
+        '[quantizeLinear] Test block-wise quantization with block_size = [3, 2, 5].',
+    input: {dataType: 'float32', shape: [6, 4, 5]},
+    scale: {dataType: 'float32', shape: [2, 2, 1]},
+    zeroPoint: {dataType: 'int8', shape: [2, 2, 1]},
+    output: {dataType: 'int8', shape: [6, 4, 5]},
+  },
+  {
+    name:
+        '[quantizeLinear] Throw if the scale size is not a factor of input size.',
     input: {dataType: 'float32', shape: [3, 2, 5]},
-    scale: {dataType: 'float32', shape: [2]},
-    zeroPoint: {dataType: 'int8', shape: [5]},
+    scale: {dataType: 'float32', shape: [2, 1, 5]},
+    zeroPoint: {dataType: 'int8', shape: [2, 1, 5]},
   },
   {
     name:
-        '[quantizeLinear] Throw if the shape of zero_point is not broadcastable to the shape of input.',
+        '[quantizeLinear] Throw if the shape of zero_point is not the same as the shape of input.',
     input: {dataType: 'float32', shape: [3, 2, 5]},
     scale: {dataType: 'float32', shape: [5]},
-    zeroPoint: {dataType: 'int8', shape: [2]},
+    zeroPoint: {dataType: 'int8', shape: [2, 5]},
   },
   {
     name:
         '[quantizeLinear] Throw if the data type of input is not the same as scale.',
     input: {dataType: 'float32', shape: [3, 2, 5]},
-    scale: {dataType: 'float16', shape: [5]},
-    zeroPoint: {dataType: 'int8', shape: [5]},
+    scale: {dataType: 'float16', shape: [3, 1, 5]},
+    zeroPoint: {dataType: 'int8', shape: [3, 1, 5]},
   },
   {
-    name: '[quantizeLinear] Throw if the data type of input is not float32.',
+    name:
+        '[quantizeLinear] Throw if the data type of input is not float32 or float16.',
     input: {dataType: 'int32', shape: [3, 2, 5]},
     scale: {dataType: 'float32', shape: [5]},
     zeroPoint: {dataType: 'int8', shape: [5]},
   },
   {
-    name: '[quantizeLinear] Throw if the data type of scale is not float32.',
+    name:
+        '[quantizeLinear] Throw if the data type of scale is not float32 or float16.',
     input: {dataType: 'float32', shape: [3, 2, 5]},
     scale: {dataType: 'int32', shape: [5]},
     zeroPoint: {dataType: 'uint8', shape: [5]},