M1 Ultra — benchmark record

M1 Ultra local LLM benchmarks across 64 GB RAM tiers on Apple Silicon Mac. 5 published rows across 2 models with explicit evidence state and RAM-tier comparison. Peak published speed is 29.7 tok/s.

5Benchmark rows

2Models tested

1RAM configurations

29.7Fastest 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 GBM1 Ultra (64 GB)5 rows · 2 models29.7 tok/s peak

All benchmark rows — M1 Ultra

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
M1 Ultra (64 GB)	Devstral Small 2 24B	4bit	—	29.7 tok/s	—	MLX	ref
M1 Ultra (64 GB)	Devstral Small 2 24B	Q4_K - Medium	—	25.3 tok/s	—	llama.cpp	ref
M1 Ultra (64 GB)	Devstral Small 2 24B	Q8	—	23.4 tok/s	—	llama.cpp	ref
M1 Ultra (64 GB)	Devstral Small 2 24B	8bit	—	22.3 tok/s	—	MLX	ref
M1 Ultra (64 GB)	Llama 3.3 70B	4bit	—	12.6 tok/s	—	LM Studio	ref

Models tested on M1 Ultra

Devstral Small 2 24B Llama 3.3 70B

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 →