479
"tHe PrOgReSs hAs bEeN iNsAnE"
this post was submitted on 11 Nov 2026
479 points (97.4% liked)
Fuck AI
8755 readers
1531 users here now
"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
AI, in this case, refers to LLMs, GPT technology, and anything listed as "AI" meant to increase market valuations.
Rules:
- Relevant posts only.
- No bigotry.
- No misinformation.
- No NSFW/NSFL content without being labeled NSFW or being in an appropriately-labeled spoiler tag.
- This isn't a place to talk about the cons and (supposed) pros of AI for different use cases. This is /c/fuck_ai .
- Especially no AI advocacy.
- No AI-generated content. If you want to post examples of AI failing hillariously, there are other communities for that like !ArtificialIgnorance@lemmy.ca .
- Venting is part of the purpose of this community, but don't aim frustration or accusations at other users.
- These rules are a living document that may change in the future.
You can help the mod team by making your reports detailed.
Effective reports:
- "The third image is a well-known anti-semitic trope"
- "Claiming that 'agentic AI' can outperform humans at all jobs is misinformation"
Less effective reports:
- "Breaks rules"
- "Misinformation"
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Could I ask what model you use locally? I’ve got a bunch of computers and vijia cards and would love to experiment with a local model. I’ve never used the subscription models, but I’d love to play around with something secure and private!
Bazzite -> LM Studio (installed via BoldBrew)
Uh what... model am i on now lol... hold on...
Qwen 3.5 9B DeepSeek v4 Flash by 'Jackrong'
But yeah generally, Qwen 3 / Qwen 3.5 type models, will work reasonably well and reasonably fast, at 4B or 2B. 9B works if you close basically everything other than LM Studio lol, or, this latest DeepSeek v4 Flash variant is considerably more performant than just standard Qwen 3.5 9B.
But yeah the way Bazzite does LM Studio via BoldBrew, it is at least theoretically reasonably containerized and isolated... not even gonna pretend I am competent enough to give any guarantees there, but also, I don't think anyone is capable of saying anything LLM related is 100% certainly correctly understood.
Oh uh, yeah, you're gonna wanna be docked/plugged in to a power source, if you want more than ...idk, 0.5 to 1.5 hrs of battery life while the LLM is actively working on something.
Thank you so much for your detailed response! My user tag I don’t remember setting for you proves correct again!
I never ended up picking up a steam deck cuz I don’t leave the house, so I’ll be using a desktop with maybe a 1070 or 2070s maybe? I’ve never tried Bazzite but I’ve got a bunch of different distros on various machines at home. I’ll throw Bazzite on my Medicat drive. Thank you so much again!
YMMV with other hardware, I genuinely have no clue rofl!
Maybe start with 2B... try to work up from there?
Maybe Nvidia cards are just... better generally, due to generally better CUDA support for LLMs than ROCm?
I have no idea!
The Deck uses an AMD APU, RDNA 2, shared LPDDR5 RAM... so... yeah i have no idea how to translate that to a more standard PC set up, in terms of performance expectations. Tooooo many variables.