Harris Notes

Home

❯

concepts

❯

Whisper Model Variants

Whisper Model Variants

May 02, 20261 min read

  • concept
  • general
  • resource

Whisper Model Variants

OpenAI’s Whisper model family ranges from tiny (~75MB) to large-v3 (~3GB), with the large-v3-turbo variant offering the best speed/accuracy trade-off for local use. Model size and quantization directly affect memory requirements on a 16GB Mac Mini M2.

Related

local-offline-transcription model-quantization apple-silicon-acceleration openai-whisper whisper-cpp whisperkit parakeet


Graph View

  • Whisper Model Variants
  • Related

Backlinks

  • Apple Silicon Acceleration
  • Cloud vs Local ASR Trade-offs
  • LLM Post-Processing
  • Local Offline Transcription
  • Model Quantization
  • OpenAI Whisper
  • Parakeet
  • WhisperKit
  • 20260502-0656-local-llm-for-audio-transcription-on-mac-mini-m2-1
  • index

Created with Quartz v4.5.2 © 2026

  • GitHub
  • Discord Community