Tether QVAC TranscribePsy: open-source 254M on-device speech recognition
Tether AI Research released TranscribePsy, an open-source 254M-parameter multilingual ASR model that runs entirely on-device — transcribing 10 minutes of audio in 1.65s on an iPhone 17 Pro Max — aimed at smartphones, smart glasses, and humanoid robots.
Why it matters
A crypto company's AI arm ships a genuinely competitive tiny ASR (8.02% WER on the Open ASR Leaderboard in its class, 25 European languages) with Apple Neural Engine-optimized builds — privacy-first voice for glasses and robots without cloud.
What you could build with it
A smart-glasses startup could ship TranscribePsy on-device for offline captioning and voice commands, keeping conversations off the cloud and avoiding per-minute transcription API costs.
Does it hold up?
Too early to judge — announced Oct 14 and all accuracy/speed figures (8.02% WER, 10 min in 1.65s) are Tether's own. No independent benchmarks or device-integration reports yet; the open-source release means community testing should appear within days.
Built with Tether QVAC TranscribePsy
- Tether launches offline AI speech model for smart glasses and robotscrypto.news · Independent crypto-press coverage of the Oct 14 release, noting Core ML / Apple Neural Engine optimization and fully offline operation.
- Tether releases on-device speech model for smart glasses and robotsblockchain.news · Confirms open-source weights and benchmark claims; frames the release as part of Tether's QVAC AI privacy-first push.
- Tether launches offline AI speech model for smart glasses and robotslaicrypto · Notes 25 European languages on laptops, eight on phones, and dedicated Apple Neural Engine builds.
Learn more
First spotted on article: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73ElevenLabs Eleven v4 + v4 Turbo: new emotive TTS architecture with a 100ms real-time variantElevenLabs released Eleven v4 and Eleven v4 Turbo on September 28: a new text-to-speech architecture with inline…models · JEV 0.72Mistral Large 4 (Le Chonk): 1.05T-param open-weight MoEMistral's new flagship: a 1.05T-parameter mixture-of-experts multimodal model with 49B active params and 1M-token…models · JEV 0.72VoiceStudio: open-source, fully-local ElevenLabs alternative with voice cloning, dubbing and dictation in 646 languagesVoiceStudio is an open-source, fully-local voice AI studio: voice cloning from a 3-second sample, voice design, video…models · JEV 0.71JetBrains Mellum 2.1: 12B MoE coding model trained for agentsJetBrains released Mellum 2.1, a 12B MoE model (2.5B active parameters, 128K context) under Apache 2.0, trained mainly…models · JEV 0.69
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.