Google AI Developers Forum

Why does inference with XNNPACK INT8 on TensorFlow Lite run slower with 4 threads than with 2 threads on a Raspberry Pi 5?

doccocaubai January 19, 2026, 4:05am 1

Topic		Replies	Views	Activity
Int8 tflite model performs worse than float32 model on ARM Cortex-A53 General Discussion tflite	2	184	July 2, 2025
TFlite conversion is slow on raspi4 General Discussion tflite , raspberry_pi , help_request	5	1459	December 8, 2023
Performance issues with tflite quantized models on ARM Cortex-A73 CPU General Discussion models , tflite	1	160	July 31, 2024
Model Maker TF Lite Slow Inference General Discussion models , tflite , model_maker , help_request	11	5505	September 6, 2021
Quantized version of tflite model does slower inferences on Android phones General Discussion models , android , tflite , help_request	1	1023	December 13, 2022