diff --git a/src/mono/mono/mini/cpu-arm64.mdesc b/src/mono/mono/mini/cpu-arm64.mdesc
index 0f8a3a2d234550..99f55aedb3a6ef 100644
--- a/src/mono/mono/mini/cpu-arm64.mdesc
+++ b/src/mono/mono/mini/cpu-arm64.mdesc
@@ -122,7 +122,7 @@ r8const: dest:f len:20
 label: len:0
 store_membase_imm: dest:b len:20
 store_membase_reg: dest:b src1:i len:20
-storex_membase: dest:b src1:x len:16
+storex_membase: dest:b src1:x len:20
 storei1_membase_imm: dest:b len:20
 storei1_membase_reg: dest:b src1:i len:12
 storei2_membase_imm: dest:b len:20
@@ -136,7 +136,7 @@ storei1_memindex: dest:b src1:i src2:i len:4
 storei2_memindex: dest:b src1:i src2:i len:4
 storei4_memindex: dest:b src1:i src2:i len:4
 load_membase: dest:i src1:b len:20
-loadx_membase: dest:x src1:b len:16
+loadx_membase: dest:x src1:b len:20
 loadi1_membase: dest:i src1:b len:32
 loadu1_membase: dest:i src1:b len:32
 loadi2_membase: dest:i src1:b len:32
diff --git a/src/mono/mono/mini/mini-arm64.c b/src/mono/mono/mini/mini-arm64.c
index dafaabaa5ba06f..0e3da92fd41c0d 100644
--- a/src/mono/mono/mini/mini-arm64.c
+++ b/src/mono/mono/mini/mini-arm64.c
@@ -437,6 +437,7 @@ get_vector_size_macro (MonoInst *ins)
 	g_assert (ins->klass);
 	int size = mono_class_value_size (ins->klass, NULL);
 	switch (size) {
+	case 12:
 	case 16:
 		return VREG_FULL;
 	case 8:
@@ -4064,13 +4065,21 @@ mono_arch_output_basic_block (MonoCompile *cfg, MonoBasicBlock *bb)
 		case OP_STOREX_MEMBASE:
 			if (ins->klass && mono_class_value_size (ins->klass, NULL) == 8)
 				code = emit_strfpx (code, sreg1, dreg, GTMREG_TO_INT (ins->inst_offset));
-			else
+			else if (ins->klass && mono_class_value_size (ins->klass, NULL) == 12) {
+				arm_neon_ins_e (code, SIZE_4, ARMREG_IP0, sreg1, 0, 2);
+				code = emit_strfpx (code, sreg1, dreg, GTMREG_TO_INT (ins->inst_offset));
+				code = emit_strfpw (code, ARMREG_IP0, dreg , GTMREG_TO_INT (ins->inst_offset + 8));
+			} else
 				code = emit_strfpq (code, sreg1, dreg, GTMREG_TO_INT (ins->inst_offset));
 			break;
 		case OP_LOADX_MEMBASE:
 			if (ins->klass && mono_class_value_size (ins->klass, NULL) == 8)
 				code = emit_ldrfpx (code, dreg, sreg1, GTMREG_TO_INT (ins->inst_offset));
-			else
+			else if (ins->klass && mono_class_value_size (ins->klass, NULL) == 12) {
+				code = emit_ldrfpx (code, dreg, sreg1, GTMREG_TO_INT (ins->inst_offset));
+				code = emit_ldrfpw (code, ARMREG_IP0, sreg1, GTMREG_TO_INT (ins->inst_offset + 8));
+				arm_neon_ins_e (code, SIZE_4, dreg, ARMREG_IP0, 2, 0);
+			} else
 				code = emit_ldrfpq (code, dreg, sreg1, GTMREG_TO_INT (ins->inst_offset));
 			break;
 		case OP_XMOVE:
diff --git a/src/mono/mono/mini/mini-llvm.c b/src/mono/mono/mini/mini-llvm.c
index 77e389a51cf9a2..a0b197866a0dbb 100644
--- a/src/mono/mono/mini/mini-llvm.c
+++ b/src/mono/mono/mini/mini-llvm.c
@@ -469,12 +469,12 @@ ovr_tag_from_mono_vector_class (MonoClass *klass)
 	llvm_ovr_tag_t ret = 0;
 	switch (size) {
 	case 8: ret |= INTRIN_vector64; break;
+	case 12: ret |= INTRIN_vector128; break;
 	case 16: ret |= INTRIN_vector128; break;
 	}
 
 	const char *class_name = m_class_get_name (klass);
-	if (!strcmp ("Vector2", class_name) || !strcmp ("Vector4", class_name) || !strcmp ("Quaternion", class_name) || !strcmp ("Plane", class_name)) {
-		// FIXME: Support Vector3
+	if (!strcmp ("Vector2", class_name) || !strcmp ("Vector3", class_name) || !strcmp ("Vector4", class_name) || !strcmp ("Quaternion", class_name) || !strcmp ("Plane", class_name)) {
 		return ret | INTRIN_float32;
 	}
 
@@ -507,6 +507,7 @@ ovr_tag_from_llvm_type (LLVMTypeRef type)
 		unsigned int bits = mono_llvm_get_prim_size_bits (type);
 		switch (bits) {
 		case 64: ret |= INTRIN_vector64; break;
+		case 96: ret |= INTRIN_vector128; break;
 		case 128: ret |= INTRIN_vector128; break;
 		default: g_assert_not_reached ();
 		}
@@ -4168,9 +4169,18 @@ emit_entry_bb (EmitContext *ctx, LLVMBuilderRef builder)
 		case LLVMArgVtypeByRef:
 		case LLVMArgAsFpArgs:
 		{
-			if (mini_class_is_simd (ctx->cfg, mono_class_from_mono_type_internal (ainfo->type)))
-				/* Treat these as normal values */
-				ctx->values [reg] = LLVMBuildLoad2 (builder, ctx->addresses [reg]->type, ctx->addresses [reg]->value, "simd_vtype");
+			MonoClass *klass = mono_class_from_mono_type_internal (ainfo->type);
+			if (mini_class_is_simd (ctx->cfg, klass)) {
+				LLVMValueRef loadedVector = LLVMBuildLoad2 (builder, ctx->addresses [reg]->type, ctx->addresses [reg]->value, "simd_vtype");
+
+				if (mono_class_value_size (klass, NULL) == 12) {
+					LLVMValueRef zero = LLVMConstReal (LLVMFloatType (), 0.0);
+					LLVMValueRef index = LLVMConstInt (LLVMInt32Type (), 3, 0);
+					loadedVector = LLVMBuildInsertElement (builder, loadedVector, zero, index, "insert_zero");
+				}
+
+				ctx->values [reg] = loadedVector;
+			}
 			break;
 		}
 		default:
@@ -6195,13 +6205,17 @@ process_bb (EmitContext *ctx, MonoBasicBlock *bb)
 			case LLVMArgFpStruct: {
 				LLVMTypeRef ret_type = LLVMGetReturnType (ctx->lmethod_type);
 				LLVMValueRef retval, elem;
-				gboolean is_simd = mini_class_is_simd (ctx->cfg, mono_class_from_mono_type_internal (sig->ret));
+
+				MonoClass *klass=  mono_class_from_mono_type_internal (sig->ret);
+				gboolean is_simd = mini_class_is_simd (ctx->cfg, klass);
 
 				if (is_simd) {
 					retval = LLVMConstNull(ret_type);
 
 					if (lhs) {
-						int len = LLVMGetVectorSize (LLVMTypeOf (lhs));
+						// Vector3: ret_type is Vector3, lhs is Vector3 represented as a Vector4 (three elements + zero). We need to extract only the first 3 elements from lhs.
+						int len = mono_class_value_size (klass, NULL) == 12 ? 3 : LLVMGetVectorSize (LLVMTypeOf (lhs));  
+						
 						for (int i = 0; i < len; i++) {
 							elem = LLVMBuildExtractElement (builder, lhs, const_int32 (i), "extract_elem");
 							retval = LLVMBuildInsertValue (builder, retval, elem, i, "insert_val_struct");
@@ -8301,6 +8315,11 @@ MONO_RESTORE_WARNING
 
 			src = convert (ctx, LLVMBuildAdd (builder, convert (ctx, values [ins->inst_basereg], IntPtrType ()), LLVMConstInt (IntPtrType (), ins->inst_offset, FALSE), ""), pointer_type (t));
 			values [ins->dreg] = mono_llvm_build_aligned_load (builder, t, src, "", FALSE, 1);
+			if (mono_class_value_size (ins->klass, NULL) == 12) {
+        				LLVMValueRef zero = LLVMConstReal (LLVMFloatType (), 0.0);
+        				LLVMValueRef index = LLVMConstInt (LLVMInt32Type (), 3, 0);
+        				values [ins->dreg] = LLVMBuildInsertElement (builder, values [ins->dreg], zero, index, "insert_zero");
+    		}
 			break;
 		}
 		case OP_STOREX_MEMBASE: {
diff --git a/src/mono/mono/mini/mini-runtime.c b/src/mono/mono/mini/mini-runtime.c
index 2a4a2d75c16ac7..f5212c9143948a 100644
--- a/src/mono/mono/mini/mini-runtime.c
+++ b/src/mono/mono/mini/mini-runtime.c
@@ -4489,8 +4489,7 @@ init_class (MonoClass *klass)
 
 #ifdef TARGET_ARM64
 	if (!strcmp (m_class_get_name_space (klass), "System.Numerics")) {
-		// FIXME: Support Vector3 https://github.com/dotnet/runtime/issues/81501
-		if (!strcmp (name, "Vector2") || !strcmp (name, "Vector4") || !strcmp (name, "Quaternion") || !strcmp (name, "Plane"))
+		if (!strcmp (name, "Vector2") || !strcmp (name, "Vector3") ||!strcmp (name, "Vector4") || !strcmp (name, "Quaternion") || !strcmp (name, "Plane"))
 			mono_class_set_is_simd_type (klass, TRUE);
 	}
 #endif
diff --git a/src/mono/mono/mini/mini.c b/src/mono/mono/mini/mini.c
index b6935d312c6ee8..ff424783275204 100644
--- a/src/mono/mono/mini/mini.c
+++ b/src/mono/mono/mini/mini.c
@@ -4591,6 +4591,10 @@ mini_get_simd_type_info (MonoClass *klass, guint32 *nelems)
 	} else if (!strcmp (klass_name, "Vector2")) {
 		*nelems = 2;
 		return MONO_TYPE_R4;
+	} else if (!strcmp (klass_name, "Vector3")) {
+		// For LLVM SIMD support, Vector3 is treated as a 4-element vector (three elements + zero).
+		*nelems = 4;
+		return MONO_TYPE_R4;
 	} else if (!strcmp (klass_name, "Vector`1") || !strcmp (klass_name, "Vector64`1") || !strcmp (klass_name, "Vector128`1") || !strcmp (klass_name, "Vector256`1") || !strcmp (klass_name, "Vector512`1")) {
 		MonoType *etype = mono_class_get_generic_class (klass)->context.class_inst->type_argv [0];
 		int size = mono_class_value_size (klass, NULL);
@@ -4602,3 +4606,4 @@ mini_get_simd_type_info (MonoClass *klass, guint32 *nelems)
 		return MONO_TYPE_VOID;
 	}
 }
+
diff --git a/src/mono/mono/mini/mini.h b/src/mono/mono/mini/mini.h
index cd2cad8e7dcece..c9adaf023f7e97 100644
--- a/src/mono/mono/mini/mini.h
+++ b/src/mono/mono/mini/mini.h
@@ -2998,7 +2998,7 @@ mini_class_is_simd (MonoCompile *cfg, MonoClass *klass)
 		return TRUE;
 	int size = mono_type_size (m_class_get_byval_arg (klass), NULL);
 #ifdef TARGET_ARM64
-	if (size == 8 || size == 16)
+	if (size == 8 || size == 12 || size == 16)
 		return TRUE;
 #else
 	if (size == 16)
diff --git a/src/mono/mono/mini/simd-intrinsics.c b/src/mono/mono/mini/simd-intrinsics.c
index 8955ede9b10506..60f81123512893 100644
--- a/src/mono/mono/mini/simd-intrinsics.c
+++ b/src/mono/mono/mini/simd-intrinsics.c
@@ -554,13 +554,14 @@ static MonoInst*
 emit_xequal (MonoCompile *cfg, MonoClass *klass, MonoTypeEnum element_type, MonoInst *arg1, MonoInst *arg2)
 {
 #ifdef TARGET_ARM64
+	gint32 simd_size = mono_class_value_size (klass, NULL);
 	if (!COMPILE_LLVM (cfg)) {
 		MonoInst* cmp = emit_xcompare (cfg, klass, element_type, arg1, arg2);
 		MonoInst* ret = emit_simd_ins (cfg, mono_defaults.boolean_class, OP_XEXTRACT, cmp->dreg, -1);
 		ret->inst_c0 = SIMD_EXTR_ARE_ALL_SET;
 		ret->inst_c1 = mono_class_value_size (klass, NULL);
 		return ret;
-	} else if (mono_class_value_size (klass, NULL) == 16) {
+	} else if (simd_size == 12 || simd_size == 16) {
 		return emit_simd_ins (cfg, klass, OP_XEQUAL_ARM64_V128_FAST, arg1->dreg, arg2->dreg);
 	} else {
 		return emit_simd_ins (cfg, klass, OP_XEQUAL, arg1->dreg, arg2->dreg);
@@ -649,10 +650,15 @@ emit_sum_vector (MonoCompile *cfg, MonoType *vector_type, MonoTypeEnum element_t
 	MonoClass *vector_class = mono_class_from_mono_type_internal (vector_type);
 	int vector_size = mono_class_value_size (vector_class, NULL);
 	int element_size;
-
-	// FIXME: Support Vector3
+	
 	guint32 nelems;
-	mini_get_simd_type_info (vector_class, &nelems);
+ 	mini_get_simd_type_info (vector_class, &nelems);
+
+	// Override nelems for Vector3, with actual number of elements, instead of treating it as a 4-element vector (three elements + zero).
+	const char *klass_name = m_class_get_name (vector_class); 
+	if (!strcmp (klass_name, "Vector3"))
+		nelems = 3;
+
 	element_size = vector_size / nelems;
 	gboolean has_single_element = vector_size == element_size;
 
@@ -2721,6 +2727,17 @@ emit_vector_2_3_4 (MonoCompile *cfg, MonoMethod *cmethod, MonoMethodSignature *f
 			for (int i = 1; i < fsig->param_count; ++i)
 				ins = emit_vector_insert_element (cfg, klass, ins, MONO_TYPE_R4, args [i + 1], i, FALSE);
 
+			if (len == 3) {
+				static float r4_0 = 0;
+				MonoInst *zero;
+				int zero_dreg = alloc_freg (cfg);
+				MONO_INST_NEW (cfg, zero, OP_R4CONST);
+				zero->inst_p0 = (void*)&r4_0;
+				zero->dreg = zero_dreg;
+				MONO_ADD_INS (cfg->cbb, zero);
+				ins = emit_vector_insert_element (cfg, klass, ins, MONO_TYPE_R4, zero, 3, FALSE);
+			}
+
 			ins->dreg = dreg;
 
 			if (indirect) {
@@ -5923,7 +5940,7 @@ arch_emit_simd_intrinsics (const char *class_ns, const char *class_name, MonoCom
 
 	if (!strcmp (class_ns, "System.Numerics")) {
 		// FIXME: Support Vector2 https://github.com/dotnet/runtime/issues/81501
-		if (!strcmp (class_name, "Vector2") || !strcmp (class_name, "Vector4") || 
+		if (!strcmp (class_name, "Vector2") || !strcmp (class_name, "Vector3")  || !strcmp (class_name, "Vector4") || 
 			!strcmp (class_name, "Quaternion") || !strcmp (class_name, "Plane"))
 			return emit_vector_2_3_4 (cfg, cmethod, fsig, args);
 	}