Argmax Open-Source SDK provides turn-key on-device inference for speech AI on iOS and macOS. It packages open-source models (Whisper for transcription, pyannote for speaker diarization, Qwen3-TTS for text-to-speech) with optimized CoreML conversions for Apple Silicon. The SDK includes WhisperKit and SpeakerKit for running speech recognition and speaker identification entirely on-device without cloud dependencies.