Hugging Face ships Swift packages for on-device speech and decisions, running Parakeet and Qwen3-ASR on Apple platforms
pcuenq · x · 2026-10-29
Hugging Face announced two new Swift packages extending its open-model, on-device approach beyond language models:
- AnySpeechRecognizer: wraps Apple's new SpeechAnalyzer API with swappable recognition engines, including NVIDIA's Parakeet and Alibaba's Qwen3-ASR running locally via Core ML, plus speaker diarization to track who spoke when.
- AnyDecisionModel: answers yes/no, multiple-choice, and rating-scale questions with typed results and probabilities — derived directly from model probabilities rather than generated text — using local models via MLX, Core ML, or llama.cpp, or decision models from Jev or Ollama.
Both packages pull open models from the Hugging Face Hub and run entirely on device. The argument: specialized small models are faster and cheaper than routing everything through an LLM.
More from Models
- DeepSeek V5 Names Found in Official Test File — Just Mock Data, Not a Release Teaser — AIFlow_ML · 2026-10-29
- GPT-6 Sol beats Gemini 3.8 Flash 1012–972 on Physical Agent Capability Index — qinzytech · 2026-10-29
- Mistral releases open weights for its 1-trillion-parameter Large 4 model — joe4942 · 2026-10-29
- Sam Altman: internal "hold-my-beer" channel showed new model crushing hardest research problems — jxnlco · 2026-10-29
- Sentence Transformers adds model merging: 60/40 weighted blend of two MiniLM checkpoints — tomaarsen · 2026-10-29
- Unsloth quants run 1T-param Mistral Large 4 on Mac Studio: 1-bit keeps 91% accuracy at 223GB — danielhanchen · 2026-10-29