Hi everyone, Our app (ID: 6744578596) seems to be stuck in review for 10 days now. Timeline: Jul 13, 2025 at 8:47 PM: Waiting for Review Jul 14, 2025 at 7:31 PM: In review Jul 14, 2025 at 8:16 PM: Rejected Jul 14, 2025 at 10:55 PM: Waiting for Review (Addressed rejection reason and resubmitted for review) Jul 15, 2025 at 3:47 PM: In review Jul 15, 2025 at 5:27 PM: Rejected Jul 15, 2025 at 5:43 PM: Waiting for Review (Addressed rejection reason and resubmitted for review) Jul 15, 2025 at 5:49 PM In review Jul 15, 2025 at 6:29 PM Rejected Jul 15, 2025 at 9:31 PM Waiting for Review (Addressed rejection reason and resubmitted for review) Jul 15, 2025 at 9:36 PM In review Jul 15, 2025 at 9:45 PM Rejected Jul 15, 2025 at 9:54 PM Waiting for Review (Addressed rejection reason and resubmitted for review) Jul 15, 2025 at 9:57 PM Rejected Jul 15, 2025 at 10:53 PM Waiting for Review (Addressed rejection reason and resubmitted for review) Jul 15, 2025 at 11:08 PM Rejected Jul 16, 2025 at 12:23 AM Waiting for Review (Addressed rejection reason and resubmitted for review) Jul 16, 2025 at 12:30 AM In Review Jul 16, 2025 at 6:02 PM Rejected Jul 16, 2025 at 6:21 PM Waiting for Review (Addressed rejection reason and resubmitted for review) Jul 16, 2025 at 6:25 PM In Review Stuck "In Review" for two days. Removed from review and resubmitted a build based on advice from an App developer on X. Jul 18, 2025 at 6:14 PM Waiting for Review Jul 18, 2025 at 6:19 PM In review. The status hasn't changed since then. Also worth noting that the reviewer(s?) keep insisting that our app is an OpenAI wrapper app and most of the rejections were on this point. We painstakingly train our own models and port them to CoreML to run on-device. A lot of the back and forth rejections were because of this confusion. On the other hand, I see plenty of OpenAI wrapper app developers ({Plant, Bird, Cat, Dog, Rock, etc} identifiers) on social media, bragging about how their non-novel app was approved in under a day. Seems quite unfair for the few developers like us, developing user privacy respecting, offline AI apps.

I was trying the latest coremltools-8.0b1 beta on macOS 15 Beta with the intent to try using the new stateful models api in CoreML. But the conversion would always fail with the error: /AppleInternal/Library/BuildRoots/<snip>/Library/Caches/com.apple.xbs/Sources/MetalPerformanceShadersGraph/mpsgraph/MetalPerformanceShadersGraph/Core/Files/MPSGraphExecutable.mm:162: failed assertion `Error: the minimum deployment target for macOS is 14.0.0' Here's a minimal repro, which works fine with both the stable version of coremltools (7.2) and the beta version (8.0b1) on macOS Sonoma 14.5, but fails with both versions of coremltools on macOS 15.0 Beta and Xcode 16.0 Beta. Which means that this most likely isn't an issue with coremltools, but with the native compilation toolchain. from collections import OrderedDict import coremltools as ct import numpy as np import torch import torch.nn as nn class ResidualAttentionBlock(nn.Module): def __init__(self, d_model: int, n_head: int, attn_mask: torch.Tensor = None): super().__init__() self.attn = nn.MultiheadAttention(d_model, n_head) self.ln_1 = nn.LayerNorm(d_model) self.mlp = nn.Sequential( OrderedDict( [ ("c_fc", nn.Linear(d_model, d_model * 4)), ("gelu", nn.GELU()), ("c_proj", nn.Linear(d_model * 4, d_model)), ] ) ) self.ln_2 = nn.LayerNorm(d_model) self.attn_mask = attn_mask def attention(self, x: torch.Tensor): self.attn_mask = ( self.attn_mask.to(dtype=x.dtype, device=x.device) if self.attn_mask is not None else None ) return self.attn(x, x, x, need_weights=False, attn_mask=self.attn_mask)[0] def forward(self, x: torch.Tensor): x = x + self.attention(self.ln_1(x)) x = x + self.mlp(self.ln_2(x)) return x class Transformer(nn.Module): def __init__( self, width: int, layers: int, heads: int, attn_mask: torch.Tensor = None ): super().__init__() self.width = width self.layers = layers self.resblocks = nn.Sequential( *[ResidualAttentionBlock(width, heads, attn_mask) for _ in range(layers)] ) def forward(self, x: torch.Tensor): return self.resblocks(x) transformer = Transformer(width=512, layers=12, heads=8) emb_tokens = torch.rand((1, 512)) ct_model = ct.convert( torch.jit.trace(transformer.eval(), emb_tokens), convert_to="mlprogram", minimum_deployment_target=ct.target.macOS14, inputs=[ct.TensorType(name="embIn", shape=[1, 512])], outputs=[ct.TensorType(name="embOutput", dtype=np.float32)], )