Hugging Face introduces AnySpeechRecognizer and AnyDecisionModel for Apple platforms
Hugging Face has introduced AnySpeechRecognizer and AnyDecisionModel, which its blog describes as open models for speech and decisions on Apple platforms. The post follows the company's earlier AnyLanguageModel work for running models locally on Apple hardware.
- Hugging Face introduced AnySpeechRecognizer and AnyDecisionModel, described as open models for speech and decisions on Apple platforms.
- The page lists ten Automatic Speech Recognition models, with sizes of 0.6B, 0.8B and 2B where shown.
- The release follows AnyLanguageModel, introduced on November 20, 2025 and marked at version 1.0 on October 20, 2026.
- The available source text gives no technical details, pricing or licence information.
What Hugging Face announced
The blog post is titled "Introducing AnySpeechRecognizer and AnyDecisionModel: Open Models for Speech and Decisions on Apple Platforms." From the title, the release has two parts: one aimed at speech recognition and one aimed at decision-making, both targeting Apple platforms and both described as open.
The source text available for this article contains little of the post's body. It does not explain how the two components work, which programming interfaces they expose, or what licence applies. This article therefore limits itself to what the page confirms.
Speech recognition models on the page
The page lists ten models under the Automatic Speech Recognition label as mentioned in the article. Where sizes are shown, they are small to mid-sized by current standards:
- One model at 0.8B and one at 2B, both marked as updated on Jul 22.
- Two models at 0.6B, one updated Sep 23 and one updated Aug 5.
- Several further entries, some updated 10 to 15 days before the page was captured, and one updated Jun 9, without size information in the extracted text.
The listing also shows usage figures next to some entries, for example 442k and 329k for the two largest models and 583k for one of the 0.6B models. The source does not say how these numbers were measured or how the models relate to AnySpeechRecognizer, so no conclusions should be drawn from them.
Part of a wider Apple platform effort
The page links to two earlier posts from the same blog. "Introducing AnyLanguageModel: One API for Local and Remote LLMs on Apple Platforms" is dated November 20, 2025. "AnyLanguageModel 1.0: One Year of Local Models on Apple Platforms" is dated October 20, 2026. Both carry tags such as swift, llm, mlx and coreml.
The naming suggests that AnySpeechRecognizer and AnyDecisionModel extend this approach from language models to speech and decision tasks. The source does not state this directly, so it remains an inference from the titles and tags.
What it means for developers
For developers building apps for Apple devices, the announcement signals that Hugging Face is widening its open tooling beyond text models. The title points to speech input and decision logic as new areas covered. Details on supported devices, performance, setup and licensing are not part of the source text, so developers should consult the original Hugging Face post for specifics before planning any integration.
