gpu4.pics
data via LocalScore
Apple M1
llamafile 0.9.2
8 GB unified Apple
running Llama 3.2 1B Instruct· Q4_K_M· 1.5B
40.7
tokens / sec
228 · Modest
Generation
40.7tok/s
higher is better
Prompt / prefill
615tok/s
higher is better
Time to first token
2.11 s
lower is better
Efficiency
5.09 tok/s/W
Electricity cost
$0.009/Mtok
Power draw
8 W
Apple M1 4P+4E · 8 GB RAM · Darwin · llamafile
gpu4.pics/r/ls-result-17

All benchmarks on this rig

ModelQuantGenPromptTTFTPowerPerf/W$/MtokScore
Llama 3.2 1B Instruct1.5BQ4_K_M40.76152.11 s8 W5.09 tok/s/W$0.009/Mtok228

How it stacks up — Llama 3.2 1B Instruct

Apple M1
40.7
NVIDIA L40S
355
NVIDIA RTX 3090 Ti
354
NVIDIA H100
335
NVIDIA RTX 3090
330
NVIDIA RTX A6000
315
NVIDIA A100 80GB
308