{"schema":"postcutoff/model@1","as_of":"2026-11-19T22:11:00+01:00","url":"https://postcutoff.com/m/phonon-2/","md":"https://postcutoff.com/m/phonon-2/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"phonon-2","name":"Phonon-2 (open weights, on-device ASR)","org":"Fermion Research","family":"Phonon","released":"2026-09-28","status":"current","type":"audio/speech","modality_in":["audio"],"modality_out":["text"],"open_weights":true,"model_license":"cc-by-4.0","context_window":null,"max_output":null,"knowledge_cutoff":null,"pricing":null,"price_line":"Price not published","access":[{"provider":"Hugging Face","model_id":"FermionResearch/Phonon-2","url":"https://huggingface.co/FermionResearch/Phonon-2"},{"provider":"PyPI (CLI)","model_id":"fermion-research","url":"https://pypi.org/project/fermion-research/","docs":"https://github.com/fermionresearch/phonon"},{"provider":"Docker","model_id":"ghcr.io/fermionresearch/phonon-cpu:2.0.2"},{"provider":"Mac app (Detta)","url":"https://www.fermionresearch.com/products/detta"}],"capabilities":[{"name":"~2-bit quantised Parakeet for on-device English ASR","detail":"A quantisation-aware-trained compression of NVIDIA parakeet-tdt-0.6b-v3: encoder weights at one of five learned levels (~2.1 bits), a 164 MB download vs the 2.5 GB teacher, averaging 5.21% WER on the Open ASR Leaderboard's seven English sets (teacher 4.96%, Whisper large-v3-turbo 6.58%, vendor-run numbers).","first":false,"discovered":"launch","source":"https://huggingface.co/FermionResearch/Phonon-2"},{"name":"Fast local transcription","detail":"Vendor figures: about 174x realtime on an M5 MacBook Air (an hour of audio in ~20 s), 143x on eight Zen 5 cores, 6,680x on one H100 at batch 128.","first":false,"discovered":"launch","source":"https://huggingface.co/FermionResearch/Phonon-2"}],"entry":null,"notes":"English only. Not a new architecture: a compressed NVIDIA Parakeet TDT 0.6B v3 (tokenizer and output conventions unchanged). The launch post's 'more accurate than Whisper large at 1/10 the size' refers to whisper-large-v3-turbo in the vendor's own table; benchmarks not independently reproduced. Fermion Research is a small startup (founder Manan Gupta).","verified":"2026-09-30","body_md":"Launch post: https://x.com/yoitsmanan/status/2104990913886031993 (≈37k views). Related: [nvidia-parakeet-canary](nvidia-parakeet-canary.md), [whisper-large-v3](whisper-large-v3.md).","page_url":"https://postcutoff.com/m/phonon-2/","standing":{"kind":"speech and audio","newer_count":14,"newer":[{"id":"nemotron-3-5-transcribe","name":"NVIDIA Nemotron 3.5 Transcribe (30B-A4B)","org":"NVIDIA","released":"2026-10-28","page_url":"https://postcutoff.com/m/nemotron-3-5-transcribe/"},{"id":"hy-simulmt-30b-a3b","name":"Hy-SimulMT-30B-A3B","org":"Tencent","released":"2026-10-22","page_url":"https://postcutoff.com/m/hy-simulmt-30b-a3b/"},{"id":"eleven-realtime-v1-max","name":"Eleven Realtime v1 Max / Mini","org":"ElevenLabs","released":"2026-10-20","page_url":"https://postcutoff.com/m/eleven-realtime-v1-max/"},{"id":"voxtral-mini-transcribe-3","name":"Voxtral Mini Transcribe 3 (+ Mini Transcribe Realtime 3)","org":"Mistral AI","released":"2026-10-20","page_url":"https://postcutoff.com/m/voxtral-mini-transcribe-3/"},{"id":"voxtral-small-transcribe-3","name":"Voxtral Small Transcribe 3","org":"Mistral AI","released":"2026-10-20","page_url":"https://postcutoff.com/m/voxtral-small-transcribe-3/"}],"before":null,"after":null,"weights":"public","licence":"cc-by-4.0","licence_permissive":false,"summary":"Weights are public. Licence: CC BY 4.0. See the licence for commercial terms."}}