OpenclawGateway
Region: US
Friday 16 October 2026 17:26:49 GMT
114904
4158
648
295
Music
Download
Comments
Chaucer :
The New NVIDIA NANO is under $300 and processes 70T Ops. 😁😁😁
2026-10-17 00:51:12
12
Zachary Kyle :
i can run that on a rtx 3060ti
2026-10-27 06:21:22
0
darkman1969 :
A pair of rx 9060 will get you 32GB ram for much less cash.
2026-10-27 09:56:27
0
ComradeWilly :
Or just spend $200 on an old nvidia p40 card
2026-10-17 11:00:39
14
ion Kaneda :
I use a GTX1060M 6gb yo training Qwen 3... dependa of tour code configuration.
2026-10-27 17:07:59
0
JSB MUSIC :
I’m running a local vision model Gemma 3 4b ; gpt-oss-20b (reasoning ) a wen video model and a qwen audio model all on an RTX 3080
2026-10-26 23:21:03
1
Trail Blazer :
electrical bill each month, 50$ +
2026-10-25 23:51:49
0
backnod :
Pffft. Whatever. I’m running local Ai on 20 sound blaster cards from 1989. You wish you were me.
2026-10-17 04:32:38
124
zionismis糞 :
if you have time to wait you can run qwen3.8:27b locally on 32GB RAM. i think even using a h200 wouldn't use the max out of that model.
2026-10-26 07:50:37
1
ADTCoach :
Local AI is still trash. I have 4 - 5090s and a Mac Studio w/ 256GB RAM
2026-10-17 00:38:05
65
weird_stuff_by_willy :
Coolest thing to do is not run AI. Just don’t. We don’t need it. Everyone is better off without it.
2026-10-25 12:19:05
9
C650 :
A 3090 is not 1500-2000
2026-10-20 19:43:38
8
iamendrecamon :
did this with 4060Ti 16Gb and 3060 12Gb. running strata 😁
2026-10-25 14:29:28
7
abd1554 :
3060 12gb is good enough
2026-10-17 02:05:27
17
TiX :
a 3090 can just much, much more than a 27b. Strata, maya, and free token kai
2026-10-19 20:28:23
3
GRIMz_Void :
I'm running Qwen3.8-Flash-Next Q3 on 2x Tesla V100 32GB cards ($500-$650/each) with Strata @100-120 T/s generation & 256k context... It one shots most of the tasks I throw at it. I run Claude Code on it. All in my (loud) 4U rack.
2026-10-17 07:29:48
5
MrSmallCapIPO :
How about the ram cost?
2026-10-22 14:13:07
0
napstew :
qwen 3.8 27b is a super bad model that over thinks and hallucinates, no one is using that for coding.
2026-10-25 13:02:11
0
alcany83 :
I bought an old m1 32gb Mac studio. Runs it fine
2026-10-24 08:25:05
5
Ali :
too much money for chatbot
2026-10-25 18:00:45
0
az-an :
I have a 4090 , what local AI can I run with it ?
2026-10-25 14:17:03
0
Héctor jara :
por just 1 Intel pro b70??
2026-10-26 14:56:47
2
Adriaanus :
Thats stupid, their 24gb combined does not make them equivalent to one 24gb card. The gpus have to now exchange data over much slower connections and a 850w psu guarantees none of that
2026-10-22 22:21:28
1
3y3less :
So bad advice... Will give you 15 tok/s for 700 USD assuming hou already have high poiwer PSU and you have motherboard with free PCIE x16 slot... In reality Intel Arc Pro B70 or B65 is much better deal for 1000 USD giving you 90 tok/s with much bigger context, vision and MTP4 because it has 32 GB VRAM
2026-10-26 07:18:36
2
PURE :
You can also use a gpu rack that old crypto miners used.
2026-10-24 18:18:03
1
To see more videos from user @openclawgateway, please go to the Tikwm
homepage.
