llama.cpp Metal on Apple M5 Max (MBP 16, 40-core GPU, 128 GB): llama 7B Q4_0 tg128 (discussion #4167, 8e672ef)

COMMUNITY llama-bench posted to llama.cpp discussion #4167. llama-bench -p 512 -n 128 -ngl 99, mean of the reporter's repeats, single stream, batch size 1. The number is per-request, NOT an aggregate across concurrent requests. Model is the canonical llama-2-7b GGUF (6.74B) the thread standardises on. Posted by @CSendner, comment 16400323 (2026-03-31), explicitly to confirm the 64 GB row: M5 Max, 6+12 CPU, 40 GPU, 128 GB. Posted before the 2026-08-25 harness change, on build 8e672ef (b1550), the same build the M1-M4 scoreboard uses. The pinned c1d0e7a M5 Max row already on main (pp512 3158) is a different binary and is not this number. Backend Metal, ngl 99. Within 2 percent of the 64 GB machine on every cell, which is what a bandwidth-bound 7B should do.

Same machine and harness as the other rows in this comment. Recorded mean only; the reporter's +/- is in the methodology. format_id is null for F16 because data/quantization has no F16 GGUF record.