100X.Ai
Region: US
Sunday 05 April 2026 12:41:29 GMT
123688
6251
62
834
Music
Download
Comments
James Tubbritt :
Reports say speed is 1 token on adverage every 30 seconds. Thats pretty slow.
2026-04-05 19:17:06
163
MrHappy :
In that case i can just move the llm into swap and do all calculations there, why would i need to move into vram if im bound by my disks read speed either way? Seems inefficient unless im not seeing something
2026-04-06 22:48:06
0
Andrew LeFors :
Last Github update was 2 years ago: [2024/08/20] v2.11.0: Support Qwen2.5
2026-04-06 07:13:38
17
xyzthings :
I think it is too much slow to process.
2026-04-05 14:12:34
15
Zakch kbir :
microsoft released a tecnique 1bit the llm will work on cpu and native ram with 5/10 token /s
2026-04-23 18:52:53
0
Tom :
1 token a min btw 😂
2026-04-05 20:03:44
33
Anthony Fine :
it's very slow. but in the end of the world slow is better than nothing
2026-04-06 20:12:11
5
Arthur Pinhas :
Guys it’s unusable with the token response speed it gets.
2026-04-06 06:28:30
2
heehee :
finally, glm 5.1 fp64 in my 4gb VRAM gpu
2026-04-26 22:18:14
0
INFERNEX AI :
Use infernex.ai already included in it
2026-04-05 18:41:41
8
tofu 🇨🇳 :
Gemma 4 is better
2026-04-06 08:24:51
1
Nico :
Interesting, does it use TurboQuant too?
2026-04-14 17:32:31
2
Vincent :
Load the entire model to RAM then queue to VRAM as much as possible so it becomes fast.
2026-04-15 14:06:13
1
404 :
And then you wait 13 minutes for a single hello prompt
2026-04-22 12:55:38
0
leuk :
how much info is distributed between layers, can this be distributed over a network?
2026-04-05 19:34:21
1
ozymo :
super slow, not worth
2026-04-18 16:24:27
0
Ivan Mercado :
Whats the accuracy
2026-05-20 16:32:49
0
hcxqi :
tunak tunak
2026-05-08 17:59:58
0
helli :
Erelelelelelelelelem
2026-04-15 09:32:27
0
Y ΠF-XVI :
my gpu only have 2gb vram 🗿
2026-04-15 05:38:32
0
Lume :
but first buy a Nvidia card
2026-06-04 14:28:46
0
mdennisa :
Still need to pay the overhead of gpu power cost , electric bills 😏
2026-04-18 04:03:49
0
itzhak :
Ram or Vram ?
2026-04-15 06:45:00
0
Đả Văn Tây :
4tok/s
2026-04-17 11:45:14
0
Trent Herring :
Think I’m gonna stick with Qwen
2026-04-07 02:22:21
0
To see more videos from user @100x.ai, please go to the Tikwm
homepage.
