diff --git a/types/primitives/util.go b/types/primitives/util.go
index 86cf204..6661a30 100644
--- a/types/primitives/util.go
+++ b/types/primitives/util.go
@@ -6,6 +6,26 @@ import (
 
 var hextable = "0123456789abcdef"
 
+const nonHexMarker = 100
+
+var reverseHexTable [256]byte
+
+func init() {
+	// Pre-fill the reverseHexTable
+	for i := 0; i < len(reverseHexTable); i++ {
+		switch {
+		case i >= '0' && i <= '9':
+			reverseHexTable[i] = byte(i) - '0'
+		case i >= 'a' && i <= 'f':
+			reverseHexTable[i] = byte(i) - 'a' + 10
+		case i >= 'A' && i <= 'F':
+			reverseHexTable[i] = byte(i) - 'A' + 10
+		default:
+			reverseHexTable[i] = nonHexMarker
+		}
+	}
+}
+
 func alignBytes(b []byte, length int, bigEndian bool) []byte {
 	if length < 0 {
 		if b == nil {
@@ -72,8 +92,22 @@ func hexToBytes(s string) ([]byte, error) {
 	if len(s) >= 2 && s[0] == '0' && (s[1] == 'x' || s[1] == 'X') {
 		s = s[2:]
 	}
-	if len(s)%2 == 1 {
-		s = "0" + s
+
+	if len(s)%2 == 0 {
+		return hex.DecodeString(s)
 	}
-	return hex.DecodeString(s)
+
+	dst := make([]byte, hex.DecodedLen(1+len(s)))
+
+	if reverseHexTable[s[0]] == nonHexMarker {
+		return nil, hex.InvalidByteError(s[0])
+	}
+	dst[0] = reverseHexTable[s[0]]
+
+	// Allocation-free string -> []byte conversion there (compiler will optimize it out)
+	if _, err := hex.Decode(dst[1:], []byte(s[1:])); err != nil {
+		return nil, err
+	}
+
+	return dst, nil
 }
diff --git a/types/primitives/util_test.go b/types/primitives/util_test.go
index 0430493..2b54ee0 100644
--- a/types/primitives/util_test.go
+++ b/types/primitives/util_test.go
@@ -27,3 +27,96 @@ func TestBytesToHex(t *testing.T) {
 	assert.Equal(t, "0x0000000010", string(writeDataHex(nil, []byte{0, 0, 0, 0, 16})))
 	assert.Equal(t, "0x10", string(writeQuantityHex(nil, []byte{0, 0, 0, 0, 16})))
 }
+
+func TestHexToBytes(t *testing.T) {
+	testCases := []struct {
+		name string
+		in   string
+		out  []byte
+		err  bool
+	}{{
+		name: "empty",
+		in:   "",
+		out:  []byte{},
+	}, {
+		name: "empty with prefix",
+		in:   "0x",
+		out:  []byte{},
+	}, {
+		name: "zero without prefix",
+		in:   "00",
+		out:  []byte{0x00},
+	}, {
+		name: "even length without prefix",
+		in:   "1a",
+		out:  []byte{0x1a},
+	}, {
+		name: "uppercase without prefix",
+		in:   "FC",
+		out:  []byte{0xfc},
+	}, {
+		name: "longer, without prefix",
+		in:   "0a1b2c",
+		out:  []byte{0x0a, 0x1b, 0x2c},
+	}, {
+		name: "zero with prefix",
+		in:   "0x00",
+		out:  []byte{0x00},
+	}, {
+		name: "even length with prefix",
+		in:   "0x1a",
+		out:  []byte{0x1a},
+	}, {
+		name: "uppercase with prefix",
+		in:   "0XFC",
+		out:  []byte{0xfc},
+	}, {
+		name: "zero with odd length",
+		in:   "0x0",
+		out:  []byte{0x00},
+	}, {
+		name: "odd length without prefix",
+		in:   "a",
+		out:  []byte{0x0a},
+	}, {
+		name: "uppercase with odd length",
+		in:   "0XFC1",
+		out:  []byte{0x0f, 0xc1},
+	}, {
+		name: "longer, with odd length",
+		in:   "0xa1b2c",
+		out:  []byte{0x0a, 0x1b, 0x2c},
+	}, {
+		name: "only invalid characters",
+		in:   "0xgg",
+		err:  true,
+	}, {
+		name: "only invalid characters without prefix",
+		in:   "gg",
+		err:  true,
+	}, {
+		name: "mixed invalid characters, with odd length",
+		in:   "0x1b0gg",
+		err:  true,
+	}, {
+		name: "with whitespace",
+		in:   "0x1 a",
+		err:  true,
+	}, {
+		name: "non-ascii characters",
+		in:   "0x（╯°□°）╯ ┻━┻",
+		err:  true,
+	}}
+
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			res, err := hexToBytes(tc.in)
+			if tc.err {
+				assert.Error(t, err)
+			} else {
+				assert.NoError(t, err)
+				assert.Equal(t, tc.out, res)
+			}
+		})
+	}
+}