🔍 Read the full analysis: A New Pair Of Open Models For Speech And Decisions On Apple on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Hugging Face has announced AnySpeechRecognizer and AnyDecisionModel, Swift packages for selecting among speech-recognition and decision models through shared APIs. The packages are intended to support on-device use, but the supplied announcement does not include independent performance tests, full compatibility details or release version information.
Hugging Face has announced AnySpeechRecognizer and AnyDecisionModel, two Swift packages that let app developers select speech-recognition and decision models through shared APIs. The packages are designed to support open models and on-device processing, giving developers alternatives to relying solely on Apple’s built-in speech engine or a text-generating language model. The supplied announcement does not provide a release date, version numbers or independent performance results.
AnySpeechRecognizer mirrors the structure of Apple’s SpeechAnalyzer API, with components for transcription, detection and model-asset management. The announcement says developers can change the model used to create a transcriber while leaving much of the surrounding workflow—such as analyzer setup, asset installation and reading results—in place. Apple’s Speech framework remains available as the default engine.
The package lists NVIDIA Parakeet and Alibaba Qwen3-ASR as alternative recognition engines. According to the announcement, Parakeet v3 recognizes 25 European languages. Qwen3-ASR is listed as supporting 30 languages and 22 Chinese dialects. Their reported features differ: Parakeet supports live and file input, word timing and confidence scores, while Qwen3-ASR returns a final result for segments of up to 40 seconds and does not provide those reporting features. The source characterizes Qwen as slower and more suited to files than live use.
AnyDecisionModel handles yes-or-no, multiple-choice and rating-scale questions, returning typed results and probabilities. The announcement says it derives answers from model probabilities rather than generating text. Developers can use on-device options through MLX, Core AI or llama.cpp, or access decision models from Jev or Ollama. Both packages can download open models from the Hugging Face Hub and run them on device, though actual requirements depend on the selected model and hardware.
More Model Choice in Swift Apps
The packages address a practical development issue: applications often have to be built around the interface of a particular model or provider. Shared APIs could reduce that coupling, allowing teams to change a speech engine or decision model without rewriting as much of their app’s integration logic. Hugging Face’s announcement describes adding further recognition engines by implementing the package’s model and executor protocols.
Local execution may let developers process audio or make structured decisions without sending each request to a hosted service. That can matter for privacy, connectivity and service costs, but the announcement does not establish that every model will suit every device or use case. Model size and device performance remain tradeoffs: the source lists Parakeet downloads of about 600 MB and Qwen3-ASR downloads of 0.9–2.4 GB. Those figures describe model downloads, not total memory or storage required during operation.
For developers, the main near-term value is a way to compare models behind a familiar Swift-facing interface. It does not mean the models are equivalent: language coverage, live-input support, output details, speed and accuracy differ. Teams will still need to test the specific combination they plan to ship.
Products Worth Considering
Apple AirPods Pro 3 Wireless Earbuds
DJI Mic 3 Wireless Microphone System (2 Transmitters, 1 Receiver, Charging Case)
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
As an affiliate, we earn on qualifying purchases.
Apple’s SpeechAnalyzer as the Baseline
The announcement places AnySpeechRecognizer alongside Apple’s SpeechAnalyzer, which it says Apple introduced with macOS 26 and iOS 26. That framework processes audio files or streams and returns results asynchronously; Apple’s AssetInventory manages model installation. AnySpeechRecognizer adopts a similar development pattern while adding routes to other recognition engines.
The source also describes Core ML conversions of Parakeet, Qwen3-ASR and NVIDIA Sortformer models published on the Hugging Face Hub. It says the repositories identify source checkpoints and file checksums, and that the package pins models to fixed revisions. The project further reports that Parakeet repositories include an encoder for live input that cuts analysis time by about a third, with a somewhat higher word error rate. Those are project-reported details; the supplied material does not include the methodology needed to independently assess them.
““familiar Swift APIs with a choice of model””
— Hugging Face announcement
Performance and Compatibility Still Open
The supplied announcement gives no package version numbers, publication date or device compatibility matrix. It also provides no independent tests of accuracy, latency, memory use or power consumption. Performance statements—including the report that Parakeet’s live encoder reduces analysis time by about a third while raising word error rate—are attributed to the project, and detailed test conditions are not included.
It is not clear what minimum hardware each model needs, how the packages handle all licensing and distribution requirements, or how AnyDecisionModel’s options compare on real application tasks. The excerpt also does not establish the full scope or limitations of speaker diarization support. Model download sizes alone do not answer questions about runtime memory, installation time or performance on particular devices.
Testing on Target Devices
Developers considering the packages can consult their READMEs for implementation examples and documented differences from Apple’s Speech framework. The announcement does not identify a schedule for further releases or a next formal milestone. The practical next step is to test candidate models against the app’s own audio and decision tasks, measuring accuracy, latency, download size and memory use on the intended devices.
More detailed benchmark methods, compatibility guidance and documentation on diarization would help teams judge how the options perform beyond the examples in the announcement. Until those details are available, the packages’ shared interfaces are confirmed, while their real-world performance and deployment fit remain questions for developers to evaluate.
Key Questions
What did Hugging Face announce?
It announced AnySpeechRecognizer and AnyDecisionModel, Swift packages for using speech and decision models through shared APIs.
Which speech models are listed?
The announcement names NVIDIA Parakeet and Alibaba Qwen3-ASR as recognition options alongside Apple’s Speech framework. Their listed language coverage and output features differ.
Can the models run on a device?
The packages are intended to support on-device models and offer routes through MLX, Core AI or llama.cpp. Actual device requirements and performance are not specified in the supplied material.
Are there independent performance results?
No independent benchmark results are included in the supplied announcement. Performance statements are project-reported, and detailed test conditions are not provided.
When are the packages available, and which devices are supported?
The supplied source does not give package versions, a publication date or a device compatibility matrix. Developers will need to consult package documentation for current availability and test on their target hardware.
Source: ThorstenMeyerAI.com
