diff --git a/gpu/src/Buffer.zig b/gpu/src/Buffer.zig
index a78a6a2d..34e692f7 100644
--- a/gpu/src/Buffer.zig
+++ b/gpu/src/Buffer.zig
@@ -92,7 +92,8 @@ pub const BindingType = enum(u32) {
     read_only_storage = 0x00000003,
 };
 
-pub const BindingLayout = struct {
+pub const BindingLayout = extern struct {
+    reserved: ?*anyopaque = null,
     type: BindingType,
     has_dynamic_offset: bool,
     min_binding_size: u64,
diff --git a/gpu/src/Device.zig b/gpu/src/Device.zig
index e1bbcc93..a3bb5671 100644
--- a/gpu/src/Device.zig
+++ b/gpu/src/Device.zig
@@ -16,6 +16,7 @@ const RenderPipeline = @import("RenderPipeline.zig");
 const CommandEncoder = @import("CommandEncoder.zig");
 const ComputePipeline = @import("ComputePipeline.zig");
 const BindGroup = @import("BindGroup.zig");
+const BindGroupLayout = @import("BindGroupLayout.zig");
 
 const Device = @This();
 
@@ -28,8 +29,7 @@ pub const VTable = struct {
     reference: fn (ptr: *anyopaque) void,
     release: fn (ptr: *anyopaque) void,
     createBindGroup: fn (ptr: *anyopaque, descriptor: *const BindGroup.Descriptor) BindGroup,
-    // createBindGroupLayout: fn (ptr: *anyopaque, descriptor: *const BindGroupLayout.Descriptor) BindGroupLayout,
-    // WGPU_EXPORT WGPUBindGroupLayout wgpuDeviceCreateBindGroupLayout(WGPUDevice device, WGPUBindGroupLayoutDescriptor const * descriptor);
+    createBindGroupLayout: fn (ptr: *anyopaque, descriptor: *const BindGroupLayout.Descriptor) BindGroupLayout,
     // createBuffer: fn (ptr: *anyopaque, descriptor: *const Buffer.Descriptor) Buffer,
     // WGPU_EXPORT WGPUBuffer wgpuDeviceCreateBuffer(WGPUDevice device, WGPUBufferDescriptor const * descriptor);
     createCommandEncoder: fn (ptr: *anyopaque, descriptor: ?*const CommandEncoder.Descriptor) CommandEncoder,
@@ -103,8 +103,12 @@ pub inline fn getQueue(device: Device) Queue {
     return device.vtable.getQueue(device.ptr);
 }
 
-pub inline fn createBindGroup(device: Device, descriptor: *const BindGroup.Descriptor) void {
-    device.vtable.createBindGroup(device.ptr, descriptor);
+pub inline fn createBindGroup(device: Device, descriptor: *const BindGroup.Descriptor) BindGroup {
+    return device.vtable.createBindGroup(device.ptr, descriptor);
+}
+
+pub inline fn createBindGroupLayout(device: Device, descriptor: *const BindGroupLayout.Descriptor) BindGroupLayout {
+    return device.vtable.createBindGroupLayout(device.ptr, descriptor);
 }
 
 pub inline fn createShaderModule(device: Device, descriptor: *const ShaderModule.Descriptor) ShaderModule {
@@ -163,9 +167,11 @@ pub const LostReason = enum(u32) {
 
 test {
     _ = VTable;
-    _ = getQueue;
     _ = reference;
     _ = release;
+    _ = getQueue;
+    _ = createBindGroup;
+    _ = createBindGroupLayout;
     _ = createShaderModule;
     _ = nativeCreateSwapChain;
     _ = destroy;
diff --git a/gpu/src/NativeInstance.zig b/gpu/src/NativeInstance.zig
index e3183dd9..6842e8f0 100644
--- a/gpu/src/NativeInstance.zig
+++ b/gpu/src/NativeInstance.zig
@@ -365,6 +365,48 @@ const device_vtable = Device.VTable{
             return wrapBindGroup(c.wgpuDeviceCreateBindGroup(@ptrCast(c.WGPUDevice, ptr), &desc));
         }
     }).createBindGroup,
+    .createBindGroupLayout = (struct {
+        pub fn createBindGroupLayout(ptr: *anyopaque, descriptor: *const BindGroupLayout.Descriptor) BindGroupLayout {
+            var few_entries: [16]c.WGPUBindGroupLayoutEntry = undefined;
+            const entries = if (descriptor.entries.len <= 8) blk: {
+                for (descriptor.entries) |entry, i| {
+                    few_entries[i] = c.WGPUBindGroupLayoutEntry{
+                        .nextInChain = null,
+                        .binding = entry.binding,
+                        .visibility = @enumToInt(entry.visibility),
+                        .buffer = @bitCast(c.WGPUBufferBindingLayout, entry.buffer),
+                        .sampler = @bitCast(c.WGPUSamplerBindingLayout, entry.sampler),
+                        .texture = @bitCast(c.WGPUTextureBindingLayout, entry.texture),
+                        .storageTexture = @bitCast(c.WGPUStorageTextureBindingLayout, entry.storage_texture),
+                    };
+                }
+                break :blk few_entries[0..descriptor.entries.len];
+            } else blk: {
+                const mem = std.heap.page_allocator.alloc(c.WGPUBindGroupLayoutEntry, descriptor.entries.len) catch unreachable;
+                for (descriptor.entries) |entry, i| {
+                    mem[i] = c.WGPUBindGroupLayoutEntry{
+                        .nextInChain = null,
+                        .binding = entry.binding,
+                        .visibility = @enumToInt(entry.visibility),
+                        .buffer = @bitCast(c.WGPUBufferBindingLayout, entry.buffer),
+                        .sampler = @bitCast(c.WGPUSamplerBindingLayout, entry.sampler),
+                        .texture = @bitCast(c.WGPUTextureBindingLayout, entry.texture),
+                        .storageTexture = @bitCast(c.WGPUStorageTextureBindingLayout, entry.storage_texture),
+                    };
+                }
+                break :blk mem;
+            };
+            defer if (entries.len > 8) std.heap.page_allocator.free(entries);
+
+            const desc = c.WGPUBindGroupLayoutDescriptor{
+                .nextInChain = null,
+                .label = if (descriptor.label) |l| l else null,
+                .entryCount = @intCast(u32, entries.len),
+                .entries = &entries[0],
+            };
+            return wrapBindGroupLayout(c.wgpuDeviceCreateBindGroupLayout(@ptrCast(c.WGPUDevice, ptr), &desc));
+        }
+    }).createBindGroupLayout,
     .createShaderModule = (struct {
         pub fn createShaderModule(ptr: *anyopaque, descriptor: *const ShaderModule.Descriptor) ShaderModule {
             switch (descriptor.code) {
@@ -676,7 +718,7 @@ const queue_vtable = Queue.VTable{
             }
 
             var few_commands: [16]c.WGPUCommandBuffer = undefined;
-            const commands = if (cmds.len <= 8) blk: {
+            const commands = if (cmds.len <= 16) blk: {
                 for (cmds) |cmd, i| {
                     few_commands[i] = @ptrCast(c.WGPUCommandBuffer, cmd.ptr);
                 }
@@ -688,7 +730,7 @@ const queue_vtable = Queue.VTable{
                 }
                 break :blk mem;
             };
-            defer if (cmds.len > 8) std.heap.page_allocator.free(cmds);
+            defer if (cmds.len > 16) std.heap.page_allocator.free(cmds);
 
             c.wgpuQueueSubmit(
                 wgpu_queue,
diff --git a/gpu/src/Sampler.zig b/gpu/src/Sampler.zig
index 061e8628..cb7bd433 100644
--- a/gpu/src/Sampler.zig
+++ b/gpu/src/Sampler.zig
@@ -34,7 +34,8 @@ pub const BindingType = enum(u32) {
     comparison = 0x00000003,
 };
 
-pub const BindingLayout = struct {
+pub const BindingLayout = extern struct {
+    reserved: ?*anyopaque = null,
     type: BindingType,
 };
 
diff --git a/gpu/src/Texture.zig b/gpu/src/Texture.zig
index 3bd33629..1bfc29c8 100644
--- a/gpu/src/Texture.zig
+++ b/gpu/src/Texture.zig
@@ -187,7 +187,8 @@ pub const SampleType = enum(u32) {
     uint = 0x00000005,
 };
 
-pub const BindingLayout = struct {
+pub const BindingLayout = extern struct {
+    reserved: ?*anyopaque = null,
     sample_type: SampleType,
     view_dimension: TextureView.Dimension,
     multisampled: bool,
diff --git a/gpu/src/structs.zig b/gpu/src/structs.zig
index 6cd68dda..454cb69e 100644
--- a/gpu/src/structs.zig
+++ b/gpu/src/structs.zig
@@ -35,7 +35,8 @@ pub const PrimitiveState = struct {
     cull_mode: CullMode,
 };
 
-pub const StorageTextureBindingLayout = struct {
+pub const StorageTextureBindingLayout = extern struct {
+    reserved: ?*anyopaque = null,
     access: StorageTextureAccess,
     format: Texture.Format,
     view_dimension: TextureView.Dimension,