Still on M1 Pro 16" MBP. Won't upgrade without the OLED panel and matrix multiplication acceleration in Apple Silicon GPUs.
Tandem OLED would actually yield a difference in everyday computing experience.
Matrix multiplication acceleration in Apple Silicon GPUs means local LLMs will run much faster. Right now, Apple Silicon machines can run very large LLM models locally decently due to high bandwidth unified memory, but the experience is often poor because processing the prompts is very slow. IE., you could be waiting minutes before the AI starts to return tokens if your context is high and the model size is large.
It doesn't make a whole lot of sense to invest in a 128GB VRAM Macbook Pro if the LLM model is going to take minutes to process my prompt.
Nvidia and AMD GPUs already have matrix multiplication accelerators. They're called Tensor Cores on Nvidia GPUs.