Model load error.

[otiswillie]

Every models loaded give me this error:

Only happens with 1.4.3, back then they runs fine.

[bilzard]

I also encounter the same problem with version 1.4.3 (14).

EDIT

In case the issue might relate to specific chip set, I share my environment:

• phone: iPhone XR
• chip: A12 Bionic chip
• OS: iOS 18.3

[ftshortt]

Same her.

[bilzard]

It seems llama.cpp will not support chip-set before A14/M1.
The main committer of llama.cpp says:

The only way is to implement the respective Metal kernels without using simd_ calls. It's not very difficult, but I don't plan on officially supporting it as it will increase the Metal code by a lot and I'm not convinced it will result in significant gains compared to CPU-only for these machines.

If somebody implements the kernels, we can put them in ggml-metal-intel.metal and have them build as a separate backend for Intel machines

ref: ggml-org/llama.cpp#4998 (comment)

If we use llama.cpp on apple device before A14/M1, it is likely to write a patch on our own.

[newmonte]

I'm also now experiencing this as of the latest update 1.4.3, no models load at all.

[mayef]

Same on DeepSeek-R1-0528-Qwen3-8B-GGUF