Apple M4 Pro — Benchmarks & Specs

Bottom line: how fast is the Apple M4 Pro?

SpecPicks median: 43.5 tokens/sec across 4 cited single-user runs of 7-9B dense models at 4-bit (Q4) under ollama (range 16.9–48.0 tok/s), from Local AI Master, LLMCheck, heyuan110.com, LocalLLaMA. For local LLM inference it generates 11.0 tokens/sec running llama3.1:34b at q4_K_M under ollama on a machine with 48 GB of unified memory, per heyuan110.com. Unified memory is the binding constraint for local inference: the Apple M4 Pro ships in configurations of up to 64 GB, per the manufacturer spec sheet, and a 34B model at q4_K_M needs roughly 21 GB for its weights alone (34B parameters at ~4.85 bits per weight, per llama.cpp), so it runs only on configurations with more unified memory than that. In PassMark CPU Mark it scores 38,018–38,028 pts, across two PassMark results (PassMark CPU benchmark listing, PassMark CPU benchmark listing). Across all its cores it scores 19,207–23,435 pts in Cinebench R23 Multi, across two published results (Notebookcheck CPU benchmark listing, CPU-Monkey CPU benchmark listing).

Every figure above is a row in the tables below, and each row links out to the review or public benchmark database the number was taken from; the memory ceiling and the weight-size estimate are not table rows and link to their sources inline; the SpecPicks median is computed here from the runs it names, each a row in the AI table. SpecPicks aggregates published measurements; it does not report first-party benchmark runs.