{"id":"fermionresearch-phonon-2","nom":"Phonon-2","editeur":"Fermion Research","pays":"États-Unis","poids_ouverts":true,"poids_selon":null,"licence":"CC BY 4.0","licence_selon":null,"liens":{"page":"https://www.fermionresearch.com/models/phonon-2/"},"fiche_auto":true,"page":"https://quelleia.com/voix/modeles/fermionresearch-phonon-2/","releve":"2026-10-19","en_bref":"35e sur 53 en transcription en anglais (Open ASR).","presentation":"Modèle de reconnaissance vocale pour l'anglais, quantifié à environ 2,1 bits par poids dans son encodeur, qui tient en un fichier de 164 Mo et produit un texte ponctué. Dérivé de Parakeet TDT 0.6B v3 de NVIDIA, il vise la transcription locale rapide sur Mac (MLX, Core ML), sur processeur ou sur GPU.","sortie":"2026-09-29","taille":"0,6 milliard de paramètres","langues":{"annoncees":{"annoncees":"anglais seulement","nombre":1,"francais":false,"liste":["en"]},"prises_en_charge_selon_open_tts":null,"evaluees":["en"]},"preuves":[{"appuie":["presentation"],"lu":"https://www.fermionresearch.com/models/phonon-2/ (« Its encoder stores every weight as one of five learned levels in about 2.1 bits. »), 17/10/2026"},{"appuie":["presentation"],"lu":"https://www.fermionresearch.com/models/phonon-2/ (« Phonon-2 Encoder about 2.1 bits per stored weight, 6-bit tables elsewhere 164 MB download · 177 MB on disk »), 17/10/2026"},{"appuie":["langues","presentation"],"lu":"https://www.fermionresearch.com/models/phonon-2/ (« Language English Audio input microphone or file 16 kHz Output Punctuated, capitalized text »), 17/10/2026"},{"appuie":["taille"],"lu":"https://www.fermionresearch.com/models/phonon-2/ (« Phonon-2 Fermion Research 0.60 B 164 MB »), 17/10/2026"},{"appuie":["presentation"],"lu":"https://huggingface.co/FermionResearch/Phonon-2/raw/main/README.md (« Based on parakeet-tdt-0.6b-v3 by NVIDIA; the tokenizer and output conventions (punctuation, casing, numerals) are the original's. Licence CC-BY-4.0, same as the original »), 17/10/2026"},{"appuie":["presentation"],"lu":"https://huggingface.co/FermionResearch/Phonon-2/raw/main/README.md (« The same package runs on Linux (x86-64 and Arm) and Windows CPUs »), 17/10/2026"},{"appuie":["sortie"],"lu":"https://www.fermionresearch.com/ (« Introducing Phonon-2 Research · September 29, 2026 »), 17/10/2026"}],"classements":[{"id":"transcription/open-asr-anglais","nom":"Transcription en anglais, Open ASR","rang":35,"sur":53,"valeur":6.87,"unite":"%","chiffres":{"rtfx":1967.1,"jeux":{"AMI-Cleaned":9.32,"Earnings22-Cleaned-AA-chunked":6.96,"Gigaspeech-Cleaned":8.35,"LS Clean":1.73,"LS Other":3.92,"Private (conversational)":14.14,"Private (scripted)":5.23,"SPGISpeech":3.7,"URGENT24":14.9,"Voice Arena Monsoon":4.91,"Voxpopuli-AA-Cleaned":2.46}},"voisins":["soundsgoodai-zipformer-cr-ctc-transducer-xl-290m","whisper-large-v3"],"source":{"nom":"Open ASR Leaderboard","mention":"Source : Open ASR Leaderboard, Hugging Face","url":"https://huggingface.co/spaces/hf-audio/open_asr_leaderboard","releve":"2026-10-19"}}],"noms_dans_les_sources":{"hf_audio":"FermionResearch/Phonon-2"}}