llamaperf

M4 Max 128GB vs M4 Max 64GB for local LLMs

15 reports on the M4 Max 128GB and 5 on the M4 Max 64GB

Which is faster for local LLMs?

On Qwen3.8 125B · 6B active at 3-bit, the M4 Max 128GB runs at 27.3 tokens per second (one run) and the M4 Max 64GB at 52.6 (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 about the same on both (546 GB/s on the M4 Max 128GB, 546 on the M4 Max 64GB).

M4 Max 128GB
Memory
128GB unified
Memory bandwidth
546 GB/s
FP16 compute
18.4 TFLOPS
Reports
15
M4 Max 64GB
Memory
64GB unified
Memory bandwidth
546 GB/s
FP16 compute
18.4 TFLOPS
Reports
5

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

ModelM4 Max 128GBM4 Max 64GB
Qwen3.8 125B · 6B active3-bit27.3one run52.6one 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.

ModelM4 Max 128GBM4 Max 64GB
Qwen3.8 125B · 6B active30.7 t/sQ5_K_S61.4 t/sQ2_K
Qwen3.8 27B6.7 t/sFP1613.5 t/sQ8_0
DeepSeek V4 Flash 284B · 13B active48.3 t/sQ2_Kdoesn't fit at these settings
Qwen3.6 35B · 3B active21.2 t/sFP1642.5 t/sQ8_0
Qwen3.6 27B6.7 t/sFP1613.5 t/sQ8_0
Gemma 4 26B · 4B active18.2 t/sFP1636.4 t/sQ8_0
DeepSeek V4.1 Flash 552B · 16B activedoesn't fit at these settingsdoesn't fit at these settings

Change the context, RAM or model in the calculator: M4 Max 128GB · M4 Max 64GB

Every report on each card, with its source: M4 Max 128GB · M4 Max 64GB

More comparisons