llamaperf

M4 16GB vs M6 32GB for local LLMs

34 reports on the M4 16GB and 22 on the M6 32GB

Which is faster for local LLMs?

On Qwen3.5 4B at 4-bit, the M4 16GB has 19.2 generation tokens/s (2 runs); prompt processing 352.4 tokens/s (one run); the M6 32GB has 16.2 generation tokens/s (one run); prompt processing 439.0 tokens/s (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 42% higher on the M6 32GB (120 GB/s on the M4 16GB, 171 on the M6 32GB). Across 2 shared models with generation measurements on both cards, the M4 16GB's median is higher on 1 and the M6 32GB's on 1.

M4 16GB
Memory
16GB unified
Memory bandwidth
120 GB/s
FP16 compute
4.3 TFLOPS
Reports
34
M6 32GB
Memory
32GB unified
Memory bandwidth
171 GB/s
FP16 compute
not on record
Reports
22

Measured on both

Generation and prompt processing speeds 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 bit class. Each side shows the engines and reported context behind its median. Context can mean a configured window or prompt depth, so these are community comparisons with differing settings. How to read these

ModelM4 16GBM6 32GB
Qwen3.5 4B4-bit19.22 runsPP 352.4 (one run)llama.cppcontext not stated16.2one runPP 439.0 (one run)MLXcontext not stated
Gemma 4 26B · 4B active4-bit30.9one runllama.cpp64K context52.2one runengine not statedcontext not stated

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 16GBM6 32GB
Qwen3.5 4B22.8 t/sQ8_017.7 t/sFP16
Gemma 4 26B · 4B activedoesn't fit at these settings23.2 t/sQ4_K_M
Qwen3.8 27Bdoesn't fit at these settings10.5 t/sQ4_K_S
Qwen3.8 125B · 6B activedoesn't fit at these settingsdoesn't fit at these settings
Qwen3.6 35B · 3B activedoesn't fit at these settings38.2 t/sQ3_K_M
DeepSeek V4 Flash 284B · 13B activedoesn't fit at these settingsdoesn't fit at these settings
DeepSeek V4.1 Flash 552B · 16B activedoesn't fit at these settingsdoesn't fit at these settings
Qwen3.6 27Bdoesn't fit at these settings10.5 t/sQ4_K_S

Change the context, RAM or model in the calculator: M4 16GB · M6 32GB

Every report on each card, with its source: M4 16GB · M6 32GB

More comparisons