Phonon-2 Releases 164MB Open ASR

// 1h agoMODEL RELEASE

Phonon-2 Releases 164MB Open ASR

Fermion Research released Phonon-2, an open English speech-recognition model with a 164 MB download, 5.21% average word error rate, and 174× realtime inference on an M5 MacBook Air. It runs locally across macOS, Linux, Windows, and NVIDIA GPUs. [Announcement](https://www.fermionresearch.com/research/phonon-2/)

// ANALYSIS

Phonon-2 makes local transcription dramatically more practical by pushing near-teacher accuracy into an edge-friendly package.

–Its 2.1-bit learned-weight encoding delivers strong accuracy per byte.
–The 164 MB model beats similarly sized alternatives such as Parakeet Redux on average WER.
–CLI, Docker, CPU, CUDA, and MLX support lowers deployment friction for developers.
–It is English-only and released under CC-BY-4.0, so commercial teams should review attribution obligations.
–The model derives from NVIDIA’s Parakeet TDT 0.6B v3, making its compactness more notable than architectural novelty. [Model card](https://huggingface.co/FermionResearch/Phonon-2)

// TAGS

phonon-2speechsttopen-weightsquantizationinferenceedge-ailocal-first

DISCOVERED

1h ago

2026-10-04

PUBLISHED

1h ago

2026-10-04

RELEVANCE

9/ 10

AUTHOR

AI Search

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

MODEL1h ago

Whistle Packs Multilingual Speech-to-Text into 16.9 MB.

Cactus Compute released Whistle, an open CPU speech-recognition model that transcribes seven languages, supports 30-second inputs, word timestamps, speech embeddings, and on-device processing. It shares the Needle runtime and weighs just 16.9 MB. [Cactus launch post](https://www.cactuscompute.com/blog/whistle)

RESEARCH1h ago

TactileStep Makes Humanoid Landings Softer

Tsinghua’s TactileStep adds sole-pressure feedback to Unitree G1 locomotion policies, using contact force, center of pressure, and loaded-area signals to regulate touchdown and stance. Tests across varied terrain report up to 48.8% lower touchdown force and 23.8% greater stance contact area. [Official project page](https://tactilestep.github.io/) · [Paper](https://arxiv.org/abs/2609.28959)

RESEARCH1h ago

DMAD Cuts Visual Generation to Few Steps

DMAD reframes distribution matching as adversarial classification, eliminating DMD’s separately fitted auxiliary score model. The research reports strong one-step image, four-step image, video, and audio-video generation results at substantially lower inference cost.