Apple M1 Max 32-core GPU 64GB llama.cpp cuBLAS: LLaMA-3 8B Q4_K_M pp1024 prefill tok/s
355.45 tok/s (pp1024 avg, per-request batch-1)
Fields
id
benchmarks/xd-apple-m1-max-8b-q4km-pp1024Metric
prefill_tok_s
Value
355.45
Unit
tok/s (pp1024 avg, per-request batch-1)
Model
LLaMA-3 8B Q4_K_M
Methodology
THIRD-PARTY MEASURED (XiongjieDai GPU-Benchmarks-on-LLM-Inference, RunPod rental). llama.cpp built with cuBLAS (make LLAMA_CUBLAS=1), May-2024 snapshot. Apple M1 Max 32-core GPU 64GB full offload (-ngl 10000); llama-bench pp/tg at 512/1024/4096/8192; value is the 1024-token column. Ubuntu 22.04, CUDA 12.1.1. Per-request batch-1, NOT aggregate.
Measured by
third_party
Reproducible
yes