M3 Max (GPU count not published) — benchmark record

M3 Max (GPU count not published) local LLM benchmarks across 64 GB RAM tiers on Apple Silicon Mac. 24 published rows across 2 models with explicit evidence state and RAM-tier comparison. Peak published speed is 10.4 tok/s.

24Benchmark rows

2Models tested

1RAM configurations

10.4Fastest 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.

64 GBM3 Max (GPU count not published, 64 GB)24 rows · 2 models10.4 tok/s peak

All benchmark rows — M3 Max (GPU count not published)

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 (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	10.4 tok/s	153.6 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	10.3 tok/s	152.1 tok/s	Ollama	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	10.3 tok/s	169.5 tok/s	Ollama	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	10.3 tok/s	164.8 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	10.3 tok/s	171.4 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	10.3 tok/s	163.8 tok/s	Ollama	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	10.2 tok/s	169.2 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	10.1 tok/s	168.3 tok/s	Ollama	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	10.1 tok/s	167.0 tok/s	Ollama	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	10.1 tok/s	166.8 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	9.9 tok/s	162.2 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	9.9 tok/s	161.5 tok/s	Ollama	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	9.7 tok/s	154.2 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	9.7 tok/s	153.0 tok/s	Ollama	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	9.2 tok/s	140.1 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	9.2 tok/s	139.0 tok/s	Ollama	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	8.6 tok/s	128.0 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	8.6 tok/s	127.1 tok/s	Ollama	ref
M3 Max (GPU count not published, 64 GB)	Llama 3.3 70B	Q4_K - Medium	—	8.2 tok/s	67.9 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	7.6 tok/s	111.2 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Qwen 3 32B	Q8_0	—	7.5 tok/s	111.8 tok/s	Ollama	ref
M3 Max (GPU count not published, 64 GB)	Llama 3.3 70B	Q4_K - Medium	—	7.5 tok/s	65.2 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Llama 3.3 70B	Q4_K - Medium	—	7.0 tok/s	59.5 tok/s	llama.cpp	ref
M3 Max (GPU count not published, 64 GB)	Llama 3.3 70B	Q4_K - Medium	—	6.1 tok/s	50.3 tok/s	llama.cpp	ref

Models tested on M3 Max (GPU count not published)

Llama 3.3 70B Qwen 3 32B

Data

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

Data from in-house lab measurements plus community-published benchmarks. See all chip families →