tlm

fenn

the tiny language model. 135 million parameters on a Raspberry Pi Zero 2 W. runs locally.

compressed output
hardware
boardPi Zero 2 W
cpuARM Cortex-A53
ram512 MB
display128x128 ST7735S
powerbattery UPS
model
nameSmolLM2-135M-Instruct
quantizationQ4_K_M
size101 MB
speed~10 tok/s
response16-22 sec
software
runtimellama.cpp v1.0.0
binary8 MB static
inferencetlm.py
displaylcd_display.py
osDebian 13 aarch64
body>