llamaperf
Submit

M1 Max 32GB vs M1 Ultra 128GB for local LLMs

2 reports on the M1 Max 32GB and 3 on the M1 Ultra 128GB

Which is faster for local LLMs?

On Qwen3.8 27B at 4-bit, the M1 Max 32GB runs at 15.8 tokens per second (one run) and the M1 Ultra 128GB at 25.1 (one run). Those are community reports whose engines and settings differ, so the gap is a rough sign, not a controlled measure of the two cards. Memory bandwidth, which caps how fast a card can write, is 2.0x higher on the M1 Ultra 128GB (410 GB/s on the M1 Max 32GB, 819 on the M1 Ultra 128GB).

M1 Max 32GB
Memory
32GB unified
Memory bandwidth
410 GB/s
FP16 compute
10.4 TFLOPS
Reports
2
M1 Ultra 128GB
Memory
128GB unified
Memory bandwidth
819 GB/s
FP16 compute
21 TFLOPS
Reports
3

Measured on both

Median tokens per second from plain runs (one device, one request, no speculative decoding, the whole model in the device's memory) of the same model size at the same quant level. Engines and context lengths can still differ between the runs. How to read these

ModelM1 Max 32GBM1 Ultra 128GB
Qwen3.8 27B4-bit15.8one run25.1one run

Estimated by the calculator

The quant the calculator recommends for each card and its estimated speed at a 32,768-token context with 32 GB of system RAM, from memory bandwidth and the model's shape. Where a model was also measured above, the measurement wins.

ModelM1 Max 32GBM1 Ultra 128GB
Qwen3.8 27B20.1 t/sQ4_K_S7.5 t/sFP16
Qwen3.8 125B · 6B activedoesn't fit at these settings34.0 t/sQ5_K_S
DeepSeek V4 Flash 284B · 13B activedoesn't fit at these settings53.5 t/sQ2_K
Qwen3.6 35B · 3B active73.2 t/sQ3_K_M23.5 t/sFP16
Qwen3.6 27B20.1 t/sQ4_K_S7.5 t/sFP16
GLM-5.3 320B · 18B activedoesn't fit at these settingsdoesn't fit at these settings
Gemma 4 26B · 4B active44.3 t/sQ4_K_M20.2 t/sFP16

Change the context, RAM or model in the calculator: M1 Max 32GB · M1 Ultra 128GB

Every report on each card, with its source: M1 Max 32GB · M1 Ultra 128GB

More comparisons