M5 Max — benchmark record

M5 Max local LLM benchmarks across 48 GB, 64 GB, 128 GB RAM tiers on Apple Silicon Mac. 60 published rows across 36 models with explicit evidence state and RAM-tier comparison. Peak published speed is 158.0 tok/s.

60Benchmark rows

36Models tested

3RAM configurations

158.0Fastest 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.

48 GBM5 Max (48 GB)4 rows · 2 models128.0 tok/s peak 64 GBM5 Max (64 GB)23 rows · 20 models148.0 tok/s peak 128 GBM5 Max (128 GB)33 rows · 23 models158.0 tok/s peak

All benchmark rows — M5 Max

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
M5 Max (128 GB)	Gemma 4 E2B	Q4_K - Medium	—	158.0 tok/s	—	MLX	ref
M5 Max (64 GB)	Qwen3.5-4B	Q4_K - Medium	—	148.0 tok/s	—	MLX	ref
M5 Max (64 GB)	Phi-4 Mini Instruct 3.8B	Q4_K - Medium	—	142.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Llama 3.1 8B	Q4_K - Medium	—	138.0 tok/s	—	MLX	ref
M5 Max (64 GB)	Qwen 3 4B	Q4_K - Medium	—	135.0 tok/s	—	Ollama	ref
M5 Max (64 GB)	Gemma 3 4B	Q4_K - Medium	—	132.0 tok/s	—	Ollama	ref
M5 Max (48 GB)	Qwen3.5-35B-A3B	4bit	—	128.0 tok/s	3235.0 tok/s	MLX	ref
M5 Max (128 GB)	Gemma 4 E4B	Q4_K - Medium	—	128.0 tok/s	—	MLX	ref
M5 Max (64 GB)	Mistral 7B v0.3	Q4_K - Medium	—	122.0 tok/s	—	Ollama	ref
M5 Max (64 GB)	Phi-4 Mini Instruct 3.8B	Q8_0	—	112.0 tok/s	—	MLX	ref
M5 Max (64 GB)	Qwen3.5-9B	Q4_K - Medium	—	105.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Qwen 3 8B	Q4_K - Medium	—	98.0 tok/s	—	Ollama	ref
M5 Max (64 GB)	Ministral 3 8B	Q4_K - Medium	—	98.0 tok/s	—	Ollama	ref
M5 Max (64 GB)	DeepSeek R1 Distill Llama 8B	Q4_K - Medium	—	97.0 tok/s	—	Ollama	ref
M5 Max (48 GB)	Qwen3.5-35B-A3B	Q4_K - Medium	—	89.4 tok/s	783.0 tok/s	llama.cpp	ref
M5 Max (64 GB)	Mistral 7B v0.3	Q8_0	—	88.0 tok/s	—	MLX	ref
M5 Max (64 GB)	Llama 3.1 8B	Q8_0	—	82.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Qwen3-Coder-Next	8bit	87.1 GB	79.3 tok/s	754.9 tok/s	MLX	ref
M5 Max (128 GB)	Qwen3.5-9B	Q8_0	—	78.0 tok/s	—	MLX	ref
M5 Max (64 GB)	DeepSeek R1 Distill Llama 8B	Q8_0	—	75.0 tok/s	—	MLX	ref
M5 Max (128 GB)	Qwen3-Coder-Next	8bit	88.2 GB	74.3 tok/s	1802.1 tok/s	MLX	ref
M5 Max (128 GB)	Qwen3-Coder-Next	8bit	89.7 GB	68.6 tok/s	1887.2 tok/s	MLX	ref
M5 Max (64 GB)	Gemma 3 12B	Q4_K - Medium	—	68.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Qwen3.5-122B-A10B	4bit	71.9 GB	65.9 tok/s	881.5 tok/s	MLX	ref
M5 Max (64 GB)	Phi-4 14B	Q4_K - Medium	—	62.0 tok/s	—	MLX	ref
M5 Max (64 GB)	Qwen 3 30B-A3B	Q4_K - Medium	—	62.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Qwen3.5-122B-A10B	4bit	73.8 GB	60.6 tok/s	1239.7 tok/s	MLX	ref
M5 Max (64 GB)	Qwen 3 14B	Q4_K - Medium	—	58.0 tok/s	—	Ollama	ref
M5 Max (64 GB)	Ministral 3 14B	Q4_K - Medium	—	58.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Qwen3.6-35B-A3B	Q4_K - Medium	—	55.0 tok/s	—	MLX	ref
M5 Max (128 GB)	Qwen3.5-122B-A10B	4bit	76.4 GB	54.9 tok/s	1067.8 tok/s	MLX	ref
M5 Max (64 GB)	Qwen3.5-35B-A3B	Q4_K - Medium	—	52.0 tok/s	—	MLX	ref
M5 Max (128 GB)	Gemma 4 26B-A4B	Q4_K - Medium	—	50.0 tok/s	—	MLX	ref
M5 Max (128 GB)	Qwen3-Coder-Next	8bit	92.6 GB	48.2 tok/s	1432.7 tok/s	MLX	ref
M5 Max (64 GB)	Qwen3.6-35B-A3B	Q4_K - Medium	—	48.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Qwen3.5-35B-A3B	Q4_K - Medium	—	48.0 tok/s	—	Ollama	ref
M5 Max (64 GB)	Gemma 3 27B	Q4_K - Medium	—	42.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Mistral Small 4 119B	Q4_K - Medium	—	42.0 tok/s	—	MLX	ref
M5 Max (128 GB)	Qwen 3 14B	Q8_0	—	42.0 tok/s	—	MLX	ref
M5 Max (128 GB)	Mistral Small 4 119B	Q4_K - Medium	—	38.0 tok/s	—	Ollama	ref
M5 Max (64 GB)	Nemotron Cascade 2 30B-A3B	Q4_K - Medium	—	35.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Qwen3.5-27B	4bit	—	31.6 tok/s	—	MLX	ref
M5 Max (48 GB)	Qwen3.5-27B	4bit	—	31.3 tok/s	779.0 tok/s	MLX	ref
M5 Max (128 GB)	Qwen 3 32B	Q4_K - Medium	—	28.0 tok/s	—	Ollama	ref
M5 Max (64 GB)	DeepSeek R1 Distill Qwen 32B	Q4_K - Medium	—	27.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Gemma 4 31B	Q4_K - Medium	—	26.0 tok/s	—	MLX	ref
M5 Max (128 GB)	Llama 4 Scout 17B-16E	Q4_K - Medium	—	26.0 tok/s	—	MLX	ref
M5 Max (48 GB)	Qwen3.5-27B	Q4_K - Medium	—	23.7 tok/s	171.0 tok/s	llama.cpp	ref
M5 Max (128 GB)	Llama 4 Scout 17B-16E	Q4_K - Medium	—	22.0 tok/s	—	Ollama	ref
M5 Max (64 GB)	Gemma 4 31B	Q4_K - Medium	—	22.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Gemma 3 27B	Q6_K	—	20.0 tok/s	391.0 tok/s	llama.cpp	ref
M5 Max (128 GB)	Qwen 3 235B-A22B	Q4_K - Medium	—	18.0 tok/s	—	MLX	ref
M5 Max (128 GB)	Qwen3.5-27B	Q6_K	—	16.5 tok/s	—	llama.cpp	ref
M5 Max (128 GB)	Llama 3.3 70B	Q4_K - Medium	—	15.0 tok/s	—	MLX	ref
M5 Max (128 GB)	Qwen 3 235B-A22B	Q4_K - Medium	—	15.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Qwen3.5-397B-A17B	4bit	—	13.0 tok/s	—	flash-moe	ref
M5 Max (128 GB)	Llama 3.3 70B	Q4_K - Medium	—	12.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	DeepSeek R1 Distill Llama 70B	Q4_K - Medium	—	11.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	Qwen 2.5 72B	Q4_K - Medium	—	10.0 tok/s	—	Ollama	ref
M5 Max (128 GB)	gpt-oss 120B	Q4_K - Medium	—	7.0 tok/s	—	Ollama	ref

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 →