The model is a 30B mixture-of-experts with roughly 4B active parameters, combining transcription across 40 languages, speaker attribution, translation, contextual prompting and audio understanding in a single checkpoint. It ranks first among open-weight models on the Hugging Face Open ASR English leaderboard with a 5.33% average word error rate, and tops VoiceArena's 14-system diarization comparison and David AI's multilingual diarization benchmark. Per Artificial Analysis, it hits 2.81% WER versus Voxtral Small's 2.77% with 24B parameters. NVIDIA says fine-tuning cut medical ASR WER by 45.8% on held-out speakers; weights, data and recipes are open.

AI summary of the linked source — not editorial content.

Story signal · how it spread on X · via Elfa

as of
Spread
i
Distinct X and Telegram accounts Elfa counted behind this story’s reports, summed over the thread. Not posts, not impressions.
16 accounts

first post
Leviathan +19.2 h later

NVDA reaction
i
Price change from the first post to 24 hours later (Hyperliquid). BTC over the same window is shown so a market-wide move is not mistaken for a reaction to this story. Not advice.
−2.0% 24h after post

$217.93 → $213.49
BTC −1.1% same window
Elfa measured −2.3% 1d · still measuring

Thread · 1 event
  1. First reportNvidia’s open speech models rank among the top transcription systems16 acc.
  2. Leviathan published this article

All times UTC.

Cited posts

  • @mihai673 MihaiFirst reportOn X ↗

    I fine-tuned NVIDIA's 0.6B Parakeet v3 on nothing but its own output, and it's now the best open model for long-form transcription on the @huggingface Open ASR Leaderboard, ahead of OpenAI's gpt-transcribe and Whisper large-v3. No human transcripts, no other model. Just a… https://t.co/Uh0vGYQwd8

  • @ArtificialAnlys Artificial AnalysisOn X ↗

    NVIDIA's Nemotron 3.5 Transcribe ties for #1 among open weights Speech to Text models with 2.8% WER, using just 4B active parameters Nemotron 3.5 Transcribe is NVIDIA's new open weights Speech to Text model for non-streaming transcription, with 30B total and 4B active…

  • @AGTPinsights AGTPOn X ↗

    NVIDIA just released Nemotron 3.5 Transcribe, an open-weight speech model that tops open-model leaderboards. Here's what you need to know. It's a promptable 30B mixture-of-experts model with about 4B active parameters. One checkpoint handles transcription in 40 languages,… https://t.co/kDyqC7Se49

Show 3 more cited posts
  • @wallstengine Wall St EngineOn X ↗

    NVIDIA JUST RELEASED NEMOTRON 3.5 TRANSCRIBE 40 languages. ~4.3B active parameters. Open weights.$NVDA reports #1 among open-weight models on Hugging Face’s English transcription leaderboard. Watch it track who said what, even when speakers talk over each other.

  • @MiaAI_lab MiaOn X ↗

    NVIDIA released Nemotron 3.5 Transcribe 🎙️ - 4B active params - Supports 40 languages - Ranks #1 open-weight model on huggingface - Leads on what was said & who said it. Could run perfectly on a DGX Spark. https://t.co/wEv6TngQyT

  • @noctus91 NoctusOn X ↗

    NVIDIA’s Nemotron 3.5 Transcribe combines multilingual speech recognition and speaker diarization in one open weight model, and its benchmark results are looking strong. It leads the DAI-ASR-I18N diarization benchmark with 18.26% DER, while posting 4.71% mean WER on the English… https://t.co/CFD84JKUh3

10 more accounts posted about this · 1 Telegram channel (no public link)

TradersTrader calls tracked by Elfa · not advice

Positioning · NVDA

Open421 open positions · 373 long · 48 short

Tracked accounts lean long about 97% of the time, so a long-heavy book is normal, not a signal. As of 29 Oct 01:51 UTC.

Reaction since the first post

Since 28 Oct 13:45 UTC2 calls opened · 1 long · 1 short

Callers with a scored record

@ThematicTrader short · entry $216.11alpha −0.163 · 46% hit rate · 65 scored

Alpha, hit rate and sample size exactly as Elfa reports them.

Whole equity market, not this story

Crowding11th percentile · vs last 28 days, 50 typical
Pulse 24h72/100 · unusualness of new calls, 0–100, 50 typical, higher means more long than usual
Book92.7% long · 2,906 accounts

As of 29 Oct 08:55 UTC.

Social-monitoring data, not investment advice Elfa prices Hyperliquid post text from Xas of

Explore the topic

Loading related coverage…

Explore the full feed

Comments

Loading comments…