{"rewrite":{"id":"r_f5babab4246cd21d3c394e83","clusterId":"c_d8708231543584de25f459f0","slug":"nvidia-releases-nemotron-3-5-transcribe-as-an-open-model","model":"deepseek-v4-1-flash","headline":"NVIDIA Releases Nemotron 3.5 Transcribe as an Open Model","summary":"NVIDIA released the transcription AI Nemotron 3.5 Transcribe on October 28, 2026. The model has 34 billion total parameters and 4.3 billion active parameters, handles audio files up to one hour long, and supports 40 languages including Japanese, which NVIDIA rates in its Production-ready accuracy tier. The model is free under the OpenMDW-1.1 license, and the Granary v2 training dataset is also out.","whyItMatters":"NVIDIA is shipping a free open transcription model that it rates as the top-performing open entry on the Open ASR Leaderboard's English section and on the Artificial Analysis Word Error Rate Index, which puts a 4.3B-active-parameter model in the same accuracy conversation as much heavier ones.","webCardHtml":"\u003cp\u003eThe model is a mixture-of-experts design: 34 billion total parameters but 4.3 billion active at a time. NVIDIA\u0026#39;s published graph plots active parameter count against error rate and places Nemotron 3.5 Transcribe at the fewest active parameters among models of comparable accuracy.\u003c/p\u003e\u003cp\u003eNVIDIA splits transcription accuracy into three tiers: Leading, Production-ready, and Supported. Japanese lands in Production-ready. The demo transcribes rapid overlapping dialogue from the 1940 film His Girl Friday and separates the speakers while doing it. On the Artificial Analysis Word Error Rate Index for non-streaming transcription, the model shares the top open-model rating with Voxtral Small Transcribe. A quantized version with lower memory use is planned.\u003c/p\u003e","blueskyPost":"NVIDIA released Nemotron 3.5 Transcribe on Oct 28: 34B total params, 4.3B active, up to one hour of audio, 40 languages, Japanese rated Production-ready. Free under OpenMDW-1.1, with the Granary v2 dataset also out. A quantized version is planned.","twitterPost":"NVIDIA released Nemotron 3.5 Transcribe on Oct 28. 34B total parameters, 4.3B active, one-hour audio, 40 languages, Japanese in the Production-ready tier. Free under OpenMDW-1.1. Granary v2 dataset also released; a quantized version is planned.","threadsPost":null,"newsletterBlurb":"NVIDIA released the open transcription model Nemotron 3.5 Transcribe on October 28, 2026, a mixture-of-experts design with 34 billion total parameters and 4.3 billion active. It handles audio up to an hour and 40 languages, with Japanese in NVIDIA's Production-ready accuracy tier. The model is free under OpenMDW-1.1, the Granary v2 dataset is out, and a quantized version is planned.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20261029-nvidia-nemotron-3-5-transcribe/\",\"title\":\"NVIDIA Releases Transcription AI \\\"Nemotron 3.5 Transcribe,\\\" Capable of Highly Accurate Speaker Identification Even When Many People Speak at Once\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":5296,"outputTokens":827,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1793249367,"createdAt":"2026-10-29T04:43:36.000Z","publishedAt":"2026-10-29T04:48:24.000Z","updatedAt":"2026-10-29T04:48:24.000Z"},"cluster":{"id":"c_d8708231543584de25f459f0","canonicalTitle":"NVIDIAが文字起こしAI「Nemotron 3.5 Transcribe」を公開、大勢が同時に話す場面でも高精度な話者識別が可能","representativeArticleId":"a_bb3fbdb65f3b5628237fde77","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[\"Nemotron 3.5 Transcribe\"],\"studios\":[\"NVIDIA\"],\"people\":[],\"type\":\"announcement\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-10-29T04:13:00.000Z","lastSeenAt":"2026-10-29T04:13:00.000Z","updatedAt":"2026-10-29T04:48:24.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20261029-nvidia-nemotron-3-5-transcribe/","title":"NVIDIAが文字起こしAI「Nemotron 3.5 Transcribe」を公開、大勢が同時に話す場面でも高精度な話者識別が可能"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":["Nemotron 3.5 Transcribe"],"studios":["NVIDIA"],"people":[],"type":"announcement","domain":"other","is_roundup":false},"keyFacts":["NVIDIA released Nemotron 3.5 Transcribe on October 28, 2026.","Nemotron 3.5 Transcribe has 34 billion total parameters and 4.3 billion active parameters.","The model transcribes audio files up to one hour long and supports 40 languages, with Japanese rated Production-ready.","Nemotron 3.5 Transcribe is free under the OpenMDW-1.1 license.","NVIDIA also released the Granary v2 dataset used to develop the model."],"pageLinks":{"healthSlugs":["nemotron-3-5-transcribe","nvidia"],"storySlugs":[],"entities":[{"slug":"nemotron-3-5-transcribe","supported":1,"disputed":0,"contradicted":0,"unverified":3,"overall":"green","story_count":1,"has_dossier":false},{"slug":"nvidia","supported":5,"disputed":0,"contradicted":1,"unverified":30,"overall":"red","story_count":42,"has_dossier":true}],"stories":[],"entitiesOk":true,"storiesOk":true}}