M3 Max (30-core GPU) — LLM Benchmarks

Measured LLM inference benchmarks for M3 Max (30-core GPU) across all RAM configurations (36 GB, 96 GB). 6 benchmark rows across 3 models. Compare how RAM affects throughput. Real runs, not estimates.

6Benchmark rows

3Models tested

2RAM configurations

133.0Fastest avg tok/s

RAM configurations

Each configuration differs only in unified memory. More RAM = larger models fit. Throughput is similar across RAM tiers at the same model size.

36 GBM3 Max (30-core GPU, 36 GB)3 rows · 3 models133.0 tok/s peak 96 GBM3 Max (30-core GPU, 96 GB)3 rows · 3 models132.9 tok/s peak

All benchmark rows — M3 Max (30-core GPU)

Sorted by avg tok/s descending. Click source badge to see original measurement.

Chip (RAM)	Model	Quant	RAM req.	Avg tok/s	Prompt tok/s	Runtime	Source
M3 Max (30-core GPU, 36 GB)	Llama 3.2 1B Instruct	Q4_K - Medium	—	133.0 tok/s	2553.3 tok/s	—	ref
M3 Max (30-core GPU, 96 GB)	Llama 3.2 1B Instruct	Q4_K - Medium	—	132.9 tok/s	2520.0 tok/s	—	ref
M3 Max (30-core GPU, 96 GB)	Llama 3.1 8B Instruct	Q4_K - Medium	—	37.7 tok/s	456.8 tok/s	—	ref
M3 Max (30-core GPU, 36 GB)	Llama 3.1 8B Instruct	Q4_K - Medium	—	37.5 tok/s	443.5 tok/s	—	ref
M3 Max (30-core GPU, 96 GB)	Qwen 2.5 14B Instruct	Q4_K - Medium	—	20.8 tok/s	238.8 tok/s	—	ref
M3 Max (30-core GPU, 36 GB)	Qwen 2.5 14B Instruct	Q4_K - Medium	—	19.8 tok/s	226.2 tok/s	—	ref

Models tested on M3 Max (30-core GPU)

Llama 3.1 8B Instruct Llama 3.2 1B Instruct Qwen 2.5 14B Instruct

Data

benchmarks.json — full dataset · chips.json — chip summaries · benchmarks.csv — CSV export

Data sourced from factory lab measurements and community reference runs. See all chips →