Show HN: Which Mac runs which LLM, and how fast (macyou.co)

A new tool has been unveiled that allows users to calculate which Mac can run specific large language models (LLMs) and at what speed, particularly focusing on Apple's Silicon architecture. The calculator outputs essential metrics based on two key formulas: the memory requirements for running a model and the speed of generation, calculated from memory bandwidth divided by the size of model weights. For instance, it shows that the Llama 3.3 70B model runs on the M4 Pro with 64 GB of RAM at a throughput of approximately 5.7 tokens per second, underscoring the hardware limitations when working with substantial models. This development is significant for the AI/ML community as it provides insights into local model deployment on Mac devices, aiding both enthusiasts and professionals in selecting the appropriate hardware. The tool highlights the importance of memory bandwidth and overhead costs, offering a nuanced view that many existing calculators overlook. Additionally, the availability of dedicated rentals for high-performance Macs facilitates broader access to advanced LLM capabilities, making robust AI experimentation more practical for users who may not have the means to purchase expensive hardware outright. Overall, this resource offers invaluable information for optimizing AI workflows on Apple devices, appealing to a range of developers and researchers in the field.