@whitewhoadie: 🚨 NVIDIA JUST DROPPED A 0.6B CPU-ONLY SPEECH RECOGNITION MODEL Nemotron-3.5-ASR delivers real-time streaming ASR across 40+ languages — and it runs entirely on CPU. Ships with: • Only 0.6B parameters • Real-time streaming output • 2.5x faster than official Nemo runtime (same accuracy) • Fully offline capable • Easy integration into local agent pipelines Another strong small model for on-device/local AI stacks. 👉 https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b Who’s running local speech recognition today? Drop a 🔥
600 million params. now called 0.6 billion or 0.0006 trillion parameters
2026-06-19 00:18:27
201
Seb Hirsch :
why Nvidia make CPU models???
2026-06-20 07:31:46
1
NidofDanes :
Nvidia really sucks at this stuff. 😂
2026-06-19 16:01:05
0
wesjofficial :
If we can run it locally this could be a missing piece to the puzzle 😁
2026-06-18 18:50:21
5
King Nazty413 :
yes but you need atleast 24gb if vram im wokrong on this since last week my 3080 cant handle it...
2026-06-19 09:04:13
0
Mr. K. :
i have literally just vibecoded a speech to text app for Linux and had to use elevenlabs api to get some sane results. i guess i will beg claude to use this now
2026-07-13 14:10:40
1
vinner :
Did they copy from qwen? 😆
2026-07-27 13:21:25
0
Dylan :
I put this into a language learning app for Mac. Works well! https://ndgold.com/live-linguist/release-v2
2026-06-21 00:23:35
0
⎝⎝✧𝙅ᥲ𝙨ѻп✧⎠⎠ :
but pleas 0.6b is not more tahn 2 gb its fit in almost all newer gpu so wahts the point of it to be slower
2026-06-18 18:29:27
0
The CCSI :
i like piper
2026-06-18 23:31:21
1
✨ :
How is it better than whisper
2026-06-20 04:25:32
1
drenmorina23 :
Can you tell me more about this
2026-06-18 17:55:12
0
Nahom :
Why not just use whisper
2026-06-18 21:27:08
2
Trent :
I would actually be quite interested to see how these models compare to the output of a professional human Stenographer. I don't think AI should ever replace human stenographers for transcripts, especially in court settings, but it would be interesting to see how accurate these new speech recognition models are in comparison.
2026-06-19 14:24:16
1
papasmurf9026 :
Superwhisper mogs and it’s smaller
2026-06-20 16:56:45
3
HoneyPumpkinSpice :
I’ll check it out. I use whisper today to detect languages for subtitles. If it’s faster and as reliable then I’ll switch
2026-06-19 15:33:51
2
iimmppoossiibbllee :
any demo video?
2026-06-18 21:52:27
0
Autocomputing :
Welp, guess I'll throw that in, too.
2026-07-11 08:56:08
0
Gaymudgeon :
Folks have been sleeping on the sheer volume of models Nvidia drops on the regular.
2026-07-11 02:44:48
0
Remoter @ Tenerife :
cool, will try on my m1 air
2026-06-30 19:29:55
0
Su :
i only know hoe to use opencode and ollama, im fucked xD other model type i think
2026-06-20 06:22:02
0
Pizzi📌 :
And there is a Google model wit 240m this isnt something new and super impressive
2026-06-19 11:08:37
0
catchmeifyoucanvx :
Best tiktoker ever
2026-06-19 07:14:16
0
LucasKPinheiro :
adorei a tradução por IA do tiktok, deixe todos os seus videos assim, facilita muito a compreensão 🇧🇷
2026-06-29 13:04:29
0
To see more videos from user @whitewhoadie, please go to the Tikwm
homepage.