Introducing AnySpeechRecognizer and AnyDecisionModel: open models for speech and decisions
Two Swift packages for Apple platforms: AnySpeechRecognizer keeps Apple's speech-to-text API but allows local engines such as Parakeet or Qwen3-ASR, with speaker identification. AnyDecisionModel answers yes/no, multiple-choice and rating questions with typed results and probabilities.
What the article says
On October 29, 2026, Hugging Face introduced two Swift packages for Apple platforms: AnySpeechRecognizer and AnyDecisionModel. They let developers pick between local and hosted models behind familiar APIs, following the earlier AnyLanguageModel 1.0 release.
AnySpeechRecognizer implements Apple's SpeechAnalyzer API with a selectable engine (Parakeet v3, Parakeet Ultra, Parakeet Unified, Qwen3-ASR) and adds speaker diarization for up to four speakers. AnyDecisionModel answers yes/no, multiple-choice and rating questions by returning probabilities instead of generated text, with six backends including MLX, llama.cpp and Ollama. Models are downloaded on demand from the Hub and run on device; both packages require Swift 6.2 and at least macOS 26 or iOS 26.
Read the official article AnnuIA profile
Summary written by AnnuIA in its own words from the official page of Hugging Face. The content and rights belong to the publisher: please read the original article.