rory ridgers
Region: GB
Monday 31 August 2026 16:21:44 GMT
2517
97
31
2
Music
Download
Comments
marcin_n_0 :
I own dental practices and my team uses local models for automating work that is conected with sensitive data
2026-08-31 21:26:23
4
bengod :
I have the hardware anyway, and the biggest thing for me is an local AI that will answer questions unrestricted, and it's as good as a frontier model from 6 months ago, they will ban them, as in the wrong hands its dangerous
2026-09-02 17:04:41
0
Ivan Massow :
We reduced our token use by 15k (to 5k) by running locally. But yes small users better off using frontier. Probably nearly time to move to Qwen tho - at a third of the price. It's good.
2026-09-01 06:56:34
0
TruthSerumb :
LTX 2.3 is incredible for local video generation (If your gpu can handle it)
2026-09-01 03:51:40
0
Dan :
Mine actually does useful stuff. Just a Mac Mini running simple loops queued for when I’m at work, but saves usage going through subscriptions. Too dumb and slow for most things, but I’ve got several LLM’s with different toolsets on an SSD and a “daily driver” LLM resident on Mac to manage any swaps. But you’re right, don’t throw money at RAM and processing grunt just yet… save it because the good stuff is not far off.
2026-09-02 17:54:33
0
Ren :
Capabilitywise, I think Anthropic and OpenAI are still going to lead, but I think it's good to have local models that can compete with them not only so the bigger closed-source companies don't hold a monopoly on AI, but also because it brings much-needed competition to the space. Plus, a lot of local models like Kimi and DeepSeek are way more cost-efficient for what you're getting than Anthropic, at least for now and benchmark wise. I've been using Claude Code for most of this year, since it's basically the standard right now. But if a better, cheaper model comes out that can get 90 percent of the way there at 10x less cost, without taking more than an hour to set up in VS Code, then I'm jumping ship. Also, people claiming Anthropic is a saint-tier company just haven't done some digging. Better than OpenAI, yeah but they're also destroying books for datasets and over-hyping products to make up for the $20 in, $1-out ratio they've got going right now. Like, Mythos was so hyped, then Fable dropped, and I don't know if it's just me, but it ain't all that. I forget when I'm using Fable versus Opus. And then there's the whole Opus quality drop scandal. also 2k is enough to run high end local models
2026-08-31 22:23:24
0
eh3 :
it's a privacy thing, the people who de-google themselves etc are the people who do it. you can claim they don't look at your stuff but your actively training their data and it's in their ts and cs
2026-09-01 15:22:23
1
uncredible :
I'm just using it on my already owned Laptop to learn for something I just kinda wanna do in like 10 years
2026-08-31 17:06:44
0
xeon. :
i think local models are needed to take power away from these massive companies who will definitely screw us all over in the next couple years. however, local models are just not cost efficient atm nd that isnt gna change for a long time. power to the people tho
2026-08-31 20:26:15
0
WillWilliamWill :
Nah local ai is the way to go if you can afford it
2026-09-02 15:00:29
0
Lyra&Bacchus :
3090...
2026-09-01 14:21:49
0
Lucien :
I run local models for dataset labelling and generating training data. At scale, it’s a fraction of the cost of paying per token/API call, especially when you’re processing millions of examples. The hardware is a fixed cost, then inference is basically electricity rather than an ongoing model bill.
2026-09-01 10:16:32
1
lateside :
It’s not a trend, I run a legitimate “boring business” company and 95% of the tasks can be completed by local models with the 5% being fable overseeing. Yes I could use cheaper models w openai api but it’d still cost me like a few thousands monthly. So why not host it locally? You’re also looking at now, with the progress nowadays in a year we might be able to locally host fable level model
2026-08-31 23:58:06
0
Jeremy Fisher :
Such an odd take, are you paid by Anthropic to promote them? You don’t need a 5k rig to run a local model. And once you’ve bought it, it’s yours forever.
2026-08-31 16:38:55
2
To see more videos from user @rory.ridgers, please go to the Tikwm
homepage.
