SW ABDU
Region: DE
Saturday 18 April 2026 07:13:11 GMT
177827
13542
199
1117
Music
Download
Comments
AI_Kev :
Sure, you just need $20m of equipment and a team of 100!
2026-04-18 19:32:00
2543
Azero :
Open source llm exist, qwen the goat
2026-04-20 12:08:30
127
Crypto_society :
Basically you just need 500M and you will have a great model
2026-04-19 02:20:38
326
JBtheCat :
ollama + Gemma 4. you can come back and thank me later
2026-04-19 08:55:36
35
DönermannDesVertrauens :
"doesnt our server can run LLMs?" *our server only having cpu
2026-04-20 10:18:05
118
I'mHereForTheComments :
just pull an open source like Qwen 3. about as good as Claude and only costs the hosting cost
2026-04-19 20:44:44
47
Mr_S1n1ster :
Or just use open source 😂
2026-04-20 06:15:19
8
Ad :
Can reduce by using local llm, one time initial investment
2026-04-19 06:05:00
5
kinetic_drive :
U could finetune a local model
2026-05-17 06:30:41
0
NoName :
That would be hilarious, if I had a cofounder
2026-04-18 15:04:19
9
justlurking :
"have you ever built your own llm" in a job interview
2026-04-21 01:00:26
9
Luke :
My old boss said the same thing, that he wanted his own personal AI like mine (A 240B model running on server hardware). He had an old Asus laptop made for office and got angry when a 4B model lagged.
2026-06-06 18:39:21
3
3_Dogs_And_A_Human :
https://huggingface.co/datasets/JosephStoneCellAI/SPFsmartGATE/tree/main
2026-04-19 12:39:51
5
Lil Elles🫀💜📉📈 :
Building LLM like GPTs are not the problem, the training process, data processing and preparation, Electricity, Processing power and all that is the big deal
2026-04-19 11:58:50
9
Amirul Akmal :
request from your company to buy 1Mil worth of specs. then, use llama.cpp or equivalents. 🤣
2026-04-19 03:37:19
5
OttoDerG :
I mean you get 2 quadrillian vc funding if you try
2026-04-20 13:08:55
5
Hussein :
Database to slow we need to cache everything
2026-05-27 11:14:33
2
Bahamut's Blessings 🇨🇦 :
need an expensive lab rig and lots of data and time to wait while it trains.... let alone the ML parameters you need to give it and guidelines and all sorts of shit before you can even start training. not worth it unless you've got the time, know-how, and money to build the multi-gpu rig.
2026-04-18 20:54:15
5
Tester0101 :
honestly LLM can indeed help in many situations and sites but let a human build an database not an AI
2026-04-19 19:00:50
1
Auto Ai Advantage :
DeepSeek V4 is screaming at you, it costs so less money and it's really good
2026-06-01 13:48:24
1
ihatemuffins4 :
looks at me dead in the eyes and says “we just need to hire a data scientist” might as well wave a magic wand
2026-04-20 15:21:31
5
Dijou :
He’s got very high hopes and low pockets
2026-04-19 22:25:57
4
Diego Duval :
It’s frankly never been easier. 10k on a server rack to host a Gemma 4 31 billion parameter model and hook up your backend. API costs = 0
2026-04-26 16:14:33
1
⃟ :
What’s LLM?
2026-04-18 11:48:21
1
Nidhal :
qwen 3.6 30B A3B is a beast
2026-04-20 06:27:46
2
To see more videos from user @swabdu1, please go to the Tikwm
homepage.
