Agentic WeeklyGet the weekly

Reported News

Hugging Face releases AnySpeechRecognizer and AnyDecisionModel for on-device AI on Apple platforms

On October 29, 2026 Hugging Face announced two Swift packages for Apple platforms: AnySpeechRecognizer, which mirrors Apple's SpeechAnalyzer API but lets apps swap in NVIDIA Parakeet or Qwen3-ASR on Core ML, and AnyDecisionModel, which answers yes-or-no, multiple-choice and rating questions from model probabilities instead of generated text.

By Agentic Weekly · Evidence: Reported · How we work

Changes reported
5
listed below, each from a source we read
We tested
0
we did not run it ourselves
Gaps we flag
3
things we could not verify
Sources
1
1 primary, linked below

What changed

Changes the sources report, with the evidence status of each
#What the source saysStatus
01Hugging Face's October 29, 2026 blog post announces two Swift packages that extend its AnyLanguageModel approach to speech recognition and decisions, with familiar Swift APIs and a choice of model.Reported
02AnySpeechRecognizer mirrors Apple's SpeechAnalyzer, SpeechTranscriber, DictationTranscriber, SpeechDetector and AssetInventory APIs, with Apple's Speech framework as the default engine; NVIDIA Parakeet and Qwen3-ASR can run locally on Core ML instead, and it adds speaker diarization.Reported
03AnyDecisionModel answers yes-or-no, multiple-choice and rating-scale questions with typed results, a probability for each option and a confidence value, reading answers from model probabilities rather than generating text.Reported
04AnyDecisionModel has six backends: MLX, MLX decision heads (such as Cloudflare's Clef), Core AI, llama.cpp, TypeSafe's hosted Jev, and Ollama (version 0.35.0 or later). Without package traits it has no third-party dependencies.Reported
05Both packages require Swift 6.2 or later and macOS 26 or iOS 26 or later; the post says the Parakeet runtime and the MLX backend need Apple silicon. AnySpeechRecognizer also ships a command-line tool, hear, for transcribing local files.Reported

reported by the sourcemeasured or run by usnot verified

Primary source: Hugging Face blog: Introducing AnySpeechRecognizer and AnyDecisionModel. We are the fast, independent index; the vendor’s own page is the source of record for this release.

In the source’s words

“AnyDecisionModel answers yes-or-no, multiple-choice, and rating-scale questions with typed results and probabilities.”

“Both packages can download open models from the Hugging Face Hub and run them on device.”

“If your code uses that shared API, changing the import is enough.”

Early verdict

Swift developers already using Apple's SpeechAnalyzer API can try an open on-device engine by changing one import; teams routing tickets or classifying text on device get probabilities instead of free-text answers.

Basis: Based only on the API description and requirements in Hugging Face's blog post; we have not run either package.

What we have not verified

  • We did not install or run either package.
  • We did not check the post's speech-recognition accuracy or speed figures, or its Emoji Magnet timing on an M4 Max.
  • The post lists early version numbers (AnySpeechRecognizer from 0.1.0, AnyDecisionModel from 0.6.0); we did not check API stability or release notes on GitHub.

This brief is Reported: written from the sources below, not from running the release ourselves.

Sources

Spotted an error? See our corrections policy or email the address on the contact page. Markdown version for agents: /news/hugging-face-anyspeechrecognizer-anydecisionmodel.md

Want these briefs in your Google Top Stories? Add Agentic Weekly as a preferred source on Google.