diff --git a/lib/mps/matrix.jl b/lib/mps/matrix.jl
index c971a3218..a9d18fee6 100644
--- a/lib/mps/matrix.jl
+++ b/lib/mps/matrix.jl
@@ -211,7 +211,7 @@ function encode!(cmdbuf::MTLCommandBuffer, matmul::MPSMatrixMultiplication, left
 end
 
 """
-    matMulMPS(a::MtlMatrix, b::MtlMatrix, c::MtlMatrix, alpha=1, beta=1,
+    matmul!(a::MtlMatrix, b::MtlMatrix, c::MtlMatrix, alpha=1, beta=1,
               transpose_left=false, transpose_right=false)
 A `MPSMatrixMultiplication` kernel thay computes:
 `c = alpha * op(a) * beta * op(b) + beta * C`
diff --git a/lib/mps/matrixrandom.jl b/lib/mps/matrixrandom.jl
index e4d80ae35..cca50a95a 100644
--- a/lib/mps/matrixrandom.jl
+++ b/lib/mps/matrixrandom.jl
@@ -16,7 +16,7 @@ export MPSMatrixRandomDistributionDescriptor
     @autoproperty distributionType::MPSMatrixRandomDistribution
     @autoproperty maximum::Float32 setter=setMaximum
     @autoproperty mean::Float32 setter=setMean
-    @autoproperty minimum::Float32 setter=setMimimum
+    @autoproperty minimum::Float32 setter=setMinimum
     @autoproperty standardDeviation::Float32 setter=setStandardDeviation
 end
 
diff --git a/lib/mps/ndarray.jl b/lib/mps/ndarray.jl
index d55b87600..2bdeec285 100644
--- a/lib/mps/ndarray.jl
+++ b/lib/mps/ndarray.jl
@@ -82,7 +82,7 @@ end
 end
 
 function MPSTemporaryNDArray(cmdbuf::MTLCommandBuffer, descriptor::MPSNDArrayDescriptor)
-    @objc [MPSNDTemporaryNDArray temporaryNDArrayWithCommandBuffer:cmdbuf::id{MTLCommandBuffer}
+    @objc [MPSTemporaryNDArray temporaryNDArrayWithCommandBuffer:cmdbuf::id{MTLCommandBuffer}
                                  descriptor:descriptor::id{MPSNDArrayDescriptor}]::id{MPSTemporaryNDArray}
     return obj
 end
@@ -123,7 +123,7 @@ end
         return obj
     end
 else
-    function MPSNDArray(buffer::MTLBuffer, offset::UInt, descriptor::MPSNDArrayDescriptor)
+    function MPSNDArray(_::MTLBuffer, _::UInt, _::MPSNDArrayDescriptor)
         @assert false "Creating an MPSNDArray that shares data with user-provided MTLBuffer is only supported in macOS v15+"
     end
 end
@@ -135,20 +135,18 @@ function MPSNDArray(arr::MtlArray{T,N}) where {T,N}
     return MPSNDArray(arr.data[], UInt(arr.offset), desc)
 end
 
-function Metal.MtlArray(ndarr::MPSNDArray; storage = Metal.DefaultStorageMode)
+function Metal.MtlArray(ndarr::MPSNDArray; storage = Metal.DefaultStorageMode, async = false)
     ndims = Int(ndarr.numberOfDimensions)
     arrsize = [lengthOfDimension(ndarr,i) for i in 0:ndims-1]
     T = convert(DataType, ndarr.dataType)
     arr = MtlArray{T,ndims,storage}(undef, reverse(arrsize)...)
     dev = device(arr)
 
-    cmdBuf = MTLCommandBuffer(global_queue(dev))
-
-    exportDataWithCommandBuffer(ndarr, cmdBuf, arr.data[], T, 0, collect(sizeof(T) .* reverse(strides(arr))))
-
-    commit!(cmdBuf)
-    wait_completed(cmdBuf)
+    cmdBuf = MTLCommandBuffer(global_queue(dev)) do cmdBuf
+        exportDataWithCommandBuffer(ndarr, cmdBuf, arr.data[], T, 0, collect(sizeof(T) .* reverse(strides(arr))))
+    end
 
+    async || wait_completed(cmdBuf)
     return arr
 end
 
diff --git a/lib/mtl/device.jl b/lib/mtl/device.jl
index 9297fe344..0747006e4 100644
--- a/lib/mtl/device.jl
+++ b/lib/mtl/device.jl
@@ -91,7 +91,7 @@ MTLDevice(i::Integer) = devices()[i]
 # family
 #
 
-export supports_family, is_m3, is_m2, is_m1
+export supports_family, is_m4, is_m3, is_m2, is_m1
 
 @cenum MTLGPUFamily::NSInteger begin
     MTLGPUFamilyMetal3 = 5001 # Metal 3 support
@@ -121,4 +121,7 @@ is_m1(dev::MTLDevice) = supports_family(dev, MTLGPUFamilyApple7) &&
                         !supports_family(dev, MTLGPUFamilyApple8)
 is_m2(dev::MTLDevice) = supports_family(dev, MTLGPUFamilyApple8) &&
                         !supports_family(dev, MTLGPUFamilyApple9)
-is_m3(dev::MTLDevice) = supports_family(dev, MTLGPUFamilyApple9)
+is_m3(dev::MTLDevice) = supports_family(dev, MTLGPUFamilyApple9) &&
+                        occursin("M3", String(dev.name))
+is_m4(dev::MTLDevice) = supports_family(dev, MTLGPUFamilyApple9) &&
+                        occursin("M4", String(dev.name))