schema_version: 1.0.0
report_type: speech-asr
task_type: automatic-speech-recognition
title: Whisper Medium INT8
description: >-
  INT8-quantized automatic speech recognition family for English speech-to-text, based on an
  encoder-decoder Transformer and optimized for efficient on-device use on ARM smartphones. It uses
  weight-only dynamic INT8 quantization with FP32 activations, LiteRT/TFLite execution, and KV-cache
  autoregressive decoding to reduce model size and improve transcription speed while preserving
  accuracy for clean English audio.
id: Arm/whisper-medium-int8-litert
filename: whisper-medium-int8-litert.tflite
base_model_id: openai/whisper-medium
vendor: OpenAI
base_model_url: https://huggingface.co/openai/whisper-medium
profile: Arm-Optimized
optimization:
  optimized_by: Arm
related_technologies:
  - SME2
weight_dtype: int8
quantization:
  method: PTQ-dynamic
  weight_bits: 8
  activation_bits: 8
  symmetric: true
  mode: dynamic
  weight_granularity: per-channel
model_size_mb: 743.146
parameter_count: 763857920
format: litert