Andrey Neural
Region: PT
Tuesday 10 November 2026 21:30:26 GMT
32765
1106
29
119
Music
Download
Comments
CopiBack :
20 minutes per token
2026-11-12 01:36:51
20
koscontroller :
deadass saw this video just a few minutes ago and its the exact same script
2026-11-12 16:44:03
1
爪丨匚卄卂乇ㄥ ﾌ卂匚Ҝ丂ㄖ几 :
I'll try it see what like
2026-11-13 23:08:00
0
Kristián Žuffa :
1 tok per sometimes?
2026-11-10 21:42:46
27
Salaf As Salih :
0.000001 tokens after 1000 years 😂😂
2026-11-11 23:23:54
2
ChinaMan :
Yeah I bet it’s so slow
2026-11-11 21:13:10
0
Charlunarefin :
I’ve been waiting since last year till he finish the first word/token
2026-11-11 23:43:35
0
int1 :
1 token per gta release and high energy usage. better stick to codex my g
2026-11-12 11:35:01
0
Jakob Schulz :
Nice now I can ask it to change a comment and I can go to sleep then see that it's still in reasoning because it's so slow.
2026-11-11 01:33:59
3
bodaw :
On cpu? That is great…I bet that you type hello and you may get Hi once you return from job 😃
2026-11-11 22:23:43
0
understatements1620 :
The come back of the commodore, Ai edition
2026-11-12 21:21:22
0
Cornea Valentin :
At that scale, it would probably be cheaper to buy an NVIDIA DGX Spark. I've run models smaller than 30B parameters using system RAM and managed to get around 10 tokens per second. But nowadays, building a system with enough RAM to keep a huge model fully loaded can cost more than buying a dedicated AI workstation like the DGX Spark. And even then, RAM capacity alone doesn't solve the memory-bandwidth bottleneck.
2026-11-11 08:51:03
0
ATLaS ConTenT MacHinE :
i can on mine usable
2026-11-11 08:24:46
0
Noir :
Too bad my model (NoirCore 1.2) is 114 billion micro parameters and uses a joint utilisation thinking framework (mixture of CPU, GPU, RAM and storage). Only requires 19GB of space and 1.8GB of offload memory. 🥱
2026-11-11 22:17:19
0
username101tyt :
Yh this didn’t happen. Cloud
2026-11-12 00:38:07
0
Ritchie T :
what is this nonsense 0.3 tokens a second... shite
2026-11-12 01:36:11
0
Luna Freya :
You can start promoting your app idea 💡 and hope something is done when you retire 😂
2026-11-11 21:11:18
0
FWPiano🎹 :
"Glm 5.2 beats the models you pay for" 🤡
2026-11-12 16:32:17
0
To see more videos from user @andrey.future, please go to the Tikwm
homepage.
