~/appricorn
Region: US
Sunday 20 September 2026 12:50:50 GMT
51478
2733
303
479
Music
Download
Comments
Morozan Andrei :
and its smart like my cat
2026-10-13 17:44:40
0
Elmz :
Can I run it on a Mac Mini
2026-10-13 16:30:37
0
Lawrence :
What's Bliss?
2026-10-13 17:17:17
0
Max :
A highly quantized model like this will never perform at the same level as a full precision model hosted on a GPU cluster. Keep dreaming.
2026-09-20 19:07:00
33
Marty Doyle :
Can it code?
2026-10-12 21:20:38
0
N.E.W.T :
Any day now open AI and anthropic is gonna tell the government that letting anyone have local AI is like letting them have A nuclear bomb
2026-09-20 16:28:39
79
tcoake2 :
can you dm me. I want to build a llm and am in the unique position where the cost of the system isnt a hurdle like it is for many. I was prepping a build list, but want to build the largest I can with a reasonable cash investment. for instance, use 3k or 4k series as opposed to the over inflated 5k series GPUs. I dont mind purchasing multiple, but outside a single cpu pc one would build at home for gaming im not sure the best way to go about it. I wanted to use multiple Nvidia Blackwell pricessors
2026-10-11 12:55:49
0
1objectiveReality :
So basically - the trillions spent will never ever see even a penny profit- and we were waiting for the bubble crash - this could trigger something worse than has ever happened- and all the money is borrowed and leveraged
2026-09-20 17:01:24
10
Rube :
8 gb??????
2026-10-04 20:07:37
0
boboze :
I want to learn how to do this but don’t know where to start as a non technical
2026-10-02 16:30:54
3
mayj :
local models are good but nowhere near the frontier models. both are useful for different reasons based on budget and your goals.
2026-10-02 19:54:32
3
HuettDoge007 :
Your welcome.
2026-10-13 00:25:03
0
Codyrainy1 :
what should I use with my rtx 3080 10gb vram 16gb ram ddr5
2026-10-11 10:32:23
0
SomeUser :
Imagine going back in time and taking a local llm with you
2026-09-21 15:04:16
7
Petersilie am Zahn :
It's looping, it's not good in several languages. It's still pretty good in tool use, sure, but language? Only English and Chinese are fluent.
2026-09-26 19:08:49
1
user1688427163923 :
Local AI’s are too restricted. Gemma 4 sucks
2026-09-20 15:38:22
0
BlinDeex :
98% accuracy on a single question u would just google otherwise, quantized models cant deal with tasks taking more than couple turns due to drift and error compounding since quantizing means cutting off fractions in weights
2026-09-27 01:41:17
9
Rust Belt Gold :
Soon we'll be able to put them on an old IBM 029 puch card computer 😂🤣
2026-09-22 00:31:59
15
Moon Wind 515 :
yup .. ai data centers are not profitable
2026-09-20 16:56:24
5
j :
I love qwen but bonsai doesn't stack uo at all
2026-09-21 21:16:52
8
El Sherif :
the blue background and the lighting. are you under water?
2026-09-26 14:58:23
0
starshine :
ollama is free. uses less gb than others. and outsmarts chatgpt and Gemini.
2026-09-26 19:55:24
3
popup580 :
can i run claude opus on it? sorry, i’m dumb.
2026-10-12 19:46:50
0
try leetduck.com :
Yeah but there's an intelligence tradeoff right?
2026-09-20 16:18:55
1
Ride Out Life :
int4 and int8,m have been useless to me unfortunately
2026-10-04 15:58:02
0
To see more videos from user @appricorn, please go to the Tikwm
homepage.
