Hugging Face brings open speech and decision models to Apple apps
AnySpeechRecognizer and AnyDecisionModel let Swift apps run speech recognition and typed decisions on device through open models.
AT A GLANCE
Key points
- AnySpeechRecognizer uses Apple’s speech API surface with open models such as Parakeet and Qwen3-ASR.
- AnyDecisionModel returns typed decisions and probabilities for each option.
- Both packages can download open models from the Hugging Face Hub and run them on device.
- Published measurements use one M4 Max system and should be independently validated.
What is changing?
Hugging Face community author Mattt released two open Swift packages for Apple platforms on October 29. AnySpeechRecognizer preserves the SpeechAnalyzer API surface introduced in macOS 26 and iOS 26 while running NVIDIA Parakeet and Alibaba Qwen3-ASR models locally through Core ML. The package also adds speaker diarization for as many as four speakers.
AnyDecisionModel answers yes-or-no, multiple-choice and rating questions by deriving typed results from model probabilities instead of generating text. It supports six backends, including MLX, Core AI, llama.cpp, Jev and Ollama, and includes tools for calibration and reducing option-order bias.
The packages require Swift 6.2 and macOS 26 or iOS 26. The Parakeet and Sortformer runtimes and the MLX backend require Apple silicon. The published measurements come from a single M4 Max system, so teams should retest accuracy, latency, memory use and calibration on their own audio and decision datasets.
Explore the original sources
Source date: 29 October 2026.
Published by AI Radar: 29 Oct 2026, 16:02 (Istanbul, UTC+3). Updated: 29 Oct 2026, 16:02.
FOLLOW THE TOPIC
Topics and tags
Share your discovery.
Send this story to a friend.
Reader comments
What do you think about this development? Share your experience or question. Comments are published after editorial approval.
Your name/nickname and comment become public if approved. Read: Privacy notice.
Loading comments…