Momus - Local AI Lab
Region: US
Wednesday 23 September 2026 08:26:32 GMT
17983
653
55
27
Music
Download
Comments
Moses Mawila aka Frozilla :
5K for years of top tier subscription… more like 6-8 months.
2026-09-23 09:11:55
19
Matheprof :
If you don’t need it for security reasons it’s not worth it. I tied a maxed out Mac Studio and when it comes to coding it’s just a waste time and resources to let those open weight models spend days in some loos looking for bugs where the frontier models deliver in 2h.
2026-09-23 18:12:04
0
Tomppu :
you could've just tested the glm and qwen with subscription and determinate that both are just waste of money, same goes for the new mimo 2.6 pro
2026-09-23 09:34:37
4
Battle_Kermit :
for 5K a 200$ monthly pro codex subscription would've lasted around 25 months. I personally would prefer to have the codex, simply because of the improvements they'll make in 2 years without me having to do anything. But if you prefer to have your own, more power to you.
2026-09-23 13:54:31
0
idku222q :
problem with cloud models is u run out of usage so easily u can get so much more out of local usage
2026-09-23 10:30:30
21
emretoilet :
Just pay 100 month on claude g 🤣🤣🤣
2026-09-23 14:50:55
1
Plastic Hippie :
That’s rad
2026-09-23 17:41:21
0
that_rugged450 :
Yo @ me in the update
2026-09-23 08:40:27
2
pickle gickle :
"years of top tier subscriptions" that give you like 1h of usage 3-4 times a day versus a local ai that can run 24/7 if you want it😭
2026-09-23 12:01:05
6
bebopskxkjj :
Just to be made absolutly worthless by the smallest M5 Ultra. I get the same performance from 1500$ Strix halo btw. and it is a gaming beast as well. Not worth it at all.
2026-09-23 16:03:11
0
nio :
I need to get my second spark too
2026-09-23 14:50:01
1
Zeke’s Stuff :
I want a Spark or Mac Studio so bad. I’d have to find a business or customer that wanted hosted ai to justify buying one though. I’m jelly
2026-09-23 13:06:03
0
mashimashimarro :
The strength is not for inference for large models. The strength is in autonomous pipelines, 24/7 agentic workloads, multi-stream inference, etc. It definitely shines when you got lots of parallelism. I currently have 4x sparks running my business, continuously and autonomously training and improving models. For agentic coding alone it makes much less sense than the alternatives. You also don’t need crazy memory for a capable model now like with Qwen3.8 flash or similar. Definitely the most understood hardware in the market right now
2026-09-23 09:00:58
3
joe66395761849 :
What do you have the qwen 28 sub agents doing? What have you found the most useful?
2026-09-23 16:59:22
0
bdidnsbwkw :
Theres no reason for local models if you don’t have a strict requirement for it (banks?). It’s not even fast nor good
2026-09-23 13:05:37
0
DISTANCE HAZEL :
Two weeks and I get the easiest bag of my life. 🤪💸💰
2026-09-23 16:00:12
0
Dan :
Would it be worth it if I just wanted to code with it like I’m just like coding abs and websites I have tried Claude plans and open AI plans but I run out of usage because the models that they allow you to use are just so expensive
2026-09-23 13:24:16
0
matt_answers :
I STILL love MicroCenter and I left the US ages ago! [Fire][Fire][Fire]
2026-09-23 14:39:30
0
bdhjsrh61 :
Super impressive
2026-09-23 10:48:25
1
gaara :
do what you want, it's your money and you like it man
2026-09-23 11:59:56
0
RuxTier :
I love your mindset, You want freedom and that's the cost of it. but for me I'm beating the shit out of my rx 9070 xt 😁😂 I want to buy one of these but it expensive
2026-09-23 14:24:46
0
ruslanmulyk :
Test DS 4.1 - it is realy good.
2026-09-23 12:11:11
0
Bree :
Love it!
2026-09-23 11:02:01
0
josharmendariz455 :
Worst possible account to find me as I debate getting a 2nd every day.
2026-09-23 14:00:32
2
mars :
yeah you have 2 weeks, first day wasted
2026-09-23 16:45:47
0
To see more videos from user @momus.ai, please go to the Tikwm
homepage.
