{"id":"stepaudio-2-5-asr","nom":"StepAudio 2.5 ASR","editeur":"StepFun","pays":"Chine","poids_ouverts":false,"poids_selon":"Artificial Analysis","licence":null,"licence_selon":null,"liens":{"page":"https://platform.stepfun.ai/docs/en/guides/models/stepaudio-2.5-asr","essayer":"https://platform.stepfun.ai/docs/en/api-reference/audio/asr-sse"},"fiche_auto":true,"page":"https://quelleia.com/voix/modeles/stepaudio-2-5-asr/","releve":"2026-10-19","en_bref":"39e sur 51 en transcription en anglais (Artificial Analysis).","presentation":"Modèle de reconnaissance vocale de StepFun qui prédit plusieurs jetons à chaque étape (Multi-Token Prediction) pour réduire fortement la latence. Il transcrit le chinois et l'anglais, en envoi unique ou en flux bidirectionnel temps réel avec horodatage.","taille":"4 milliards de paramètres","langues":{"annoncees":{"annoncees":"chinois et anglais","nombre":2,"francais":false,"liste":["zh","en"]},"prises_en_charge_selon_open_tts":null,"evaluees":["en"]},"preuves":[{"appuie":["presentation","taille"],"lu":"https://platform.stepfun.ai/docs/en/guides/models/stepaudio-2.5-asr.md (« StepAudio 2.5 ASR is a 4B-parameter speech recognition model. It introduces Multi-Token Prediction (MTP) technology to predict multiple tokens per step in parallel »), 17/10/2026"},{"appuie":["langues","presentation"],"lu":"https://platform.stepfun.ai/docs/en/guides/models/audio.md (« Supports Chinese and English recognition with ITN text normalization. »), 17/10/2026"},{"appuie":["presentation"],"lu":"https://platform.stepfun.ai/docs/en/guides/models/audio.md (« stream audio over WebSocket and receive incremental and final results with per-word timing; supports server-side VAD. »), 17/10/2026"}],"classements":[{"id":"transcription/artificial-analysis","nom":"Transcription en anglais, indice AA-WER","rang":39,"sur":51,"valeur":4.67,"unite":"%","chiffres":{"prix_pour_1000_minutes":0.366667,"jeux":{"AA-AgentTalk":4.03,"Earnings22":8.12,"VoxPopuli":2.5},"date_sortie":"2026-04-22","hebergeur":"StepFun"},"voisins":["whisper-large-v3-turbo","amazon-nova-2-pro"],"source":{"nom":"Artificial Analysis","mention":"Source : Artificial Analysis","url":"https://artificialanalysis.ai/speech-to-text/models/stepaudio","releve":"2026-10-19"}}],"noms_dans_les_sources":{"artificial_analysis":"StepAudio 2.5 ASR"}}