llamaperf

M2 Pro 32GB vs M5 Ultra 256GB for local LLMs

1 report on the M2 Pro 32GB and 12 on the M5 Ultra 256GB

Which is faster for local LLMs?

On Qwen3.8 27B at 4-bit, the M2 Pro 32GB runs at 8.6 tokens per second (one run) and the M5 Ultra 256GB at 50.0 (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 6.0x higher on the M5 Ultra 256GB (205 GB/s on the M2 Pro 32GB, 1,229 on the M5 Ultra 256GB).

M2 Pro 32GB
Memory
32GB unified
Memory bandwidth
205 GB/s
FP16 compute
6.8 TFLOPS
Reports
1
M5 Ultra 256GB
Memory
256GB unified
Memory bandwidth
1,229 GB/s
FP16 compute
not on record
Reports
12

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

ModelM2 Pro 32GBM5 Ultra 256GB
Qwen3.8 27B4-bit8.6one run50.0one 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.

ModelM2 Pro 32GBM5 Ultra 256GB
Qwen3.8 27B11.5 t/sQ4_K_S14.9 t/sFP16
Qwen3.8 125B · 6B activedoesn't fit at these settings50.8 t/sQ8_0
DeepSeek V4 Flash 284B · 13B activedoesn't fit at these settings66.9 t/sQ4_K_M
Qwen3.6 35B · 3B active41.9 t/sQ3_K_M53.5 t/sFP16
Qwen3.6 27B11.5 t/sQ4_K_S14.9 t/sFP16
Gemma 4 26B · 4B active25.4 t/sQ4_K_M40.1 t/sFP16
DeepSeek V4.1 Flash 552B · 16B activedoesn't fit at these settings89.4 t/sQ2_K

Change the context, RAM or model in the calculator: M2 Pro 32GB · M5 Ultra 256GB

Every report on each card, with its source: M2 Pro 32GB · M5 Ultra 256GB

More comparisons