M2 Ultra (60-core GPU) — LLM Benchmarks

Measured LLM inference benchmarks for M2 Ultra (60-core GPU) across all RAM configurations (64 GB, 128 GB, 192 GB). 5 benchmark rows across 3 models. Compare how RAM affects throughput. Real runs, not estimates.

5Benchmark rows

3Models tested

3RAM configurations

176.4Fastest 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 GBM2 Ultra (60-core GPU, 64 GB)3 rows · 3 models174.1 tok/s peak 128 GBM2 Ultra (60-core GPU, 128 GB)1 rows · 1 models176.4 tok/s peak 192 GBM2 Ultra (60-core GPU, 192 GB)1 rows · 1 models169.8 tok/s peak

All benchmark rows — M2 Ultra (60-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
M2 Ultra (60-core GPU, 128 GB)	Llama 3.2 1B Instruct	Q4_K - Medium	—	176.4 tok/s	3295.5 tok/s	—	ref
M2 Ultra (60-core GPU, 64 GB)	Llama 3.2 1B Instruct	Q4_K - Medium	—	174.1 tok/s	3290.2 tok/s	—	ref
M2 Ultra (60-core GPU, 192 GB)	Llama 3.2 1B Instruct	Q4_K - Medium	—	169.8 tok/s	3272.2 tok/s	—	ref
M2 Ultra (60-core GPU, 64 GB)	Llama 3.1 8B Instruct	Q4_K - Medium	—	59.5 tok/s	703.3 tok/s	—	ref
M2 Ultra (60-core GPU, 64 GB)	Qwen 2.5 14B Instruct	Q4_K - Medium	—	34.2 tok/s	381.1 tok/s	—	ref

Models tested on M2 Ultra (60-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 →