M2 Ultra (76-core GPU) — LLM Benchmarks

Measured LLM inference benchmarks for M2 Ultra (76-core GPU) across all RAM configurations (128 GB, 192 GB). 2 benchmark rows across 2 models. Compare how RAM affects throughput. Real runs, not estimates.

2Benchmark rows

2Models tested

2RAM configurations

94.3Fastest 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.

128 GBM2 Ultra (76-core GPU, 128 GB)1 rows · 1 models36.6 tok/s peak 192 GBM2 Ultra (76-core GPU, 192 GB)1 rows · 1 models94.3 tok/s peak

All benchmark rows — M2 Ultra (76-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 (76-core GPU, 192 GB)	Llama 2 7B	Q4_0	3.6 GB	94.3 tok/s	1238.5 tok/s	llama.cpp	ref
M2 Ultra (76-core GPU, 128 GB)	Qwen 2.5 14B Instruct	Q4_K - Medium	—	36.6 tok/s	470.5 tok/s	—	ref

Models tested on M2 Ultra (76-core GPU)

Llama 2 7B 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 →