What AI models do you usually use most?
Displaying poll results.22716 total votes.
Most Votes
- How much of your coding is done by AI coding agents these days? Posted on June 21st, 2026 | 21950 votes
Most Comments
- How much of your coding is done by AI coding agents these days? Posted on June 21st, 2026 | 33 comments
Flux Schnell (Score:2)
Flux Schnell is the one I use most. Followed by Stabe Diffusion and DALL-E 3.
Re: (Score:1)
My human intuition said you're pulling a fast one, possibly with AIs from books or movies. Or perhaps it's a joke.
So I fed your post to an AI via the google. I think that AI Overview means Gemini? It took you completely seriously, but the AI answer did not convince me that any of them are actual AI systems. The later hits actually indicated that they probably exist in the subcategory of AI-powered image generators. (Related argument recently with someone about not trusting the AI Overview. I sure wanted to
Re: (Score:2)
My human intuition said you're pulling a fast one, possibly with AIs from books or movies. Or perhaps it's a joke.
You wrote that long ass post, didn't know WTF GP was referring to, and didn't look up any of the key terms directly?
Flux Schnell: https://huggingface.co/black-f... [huggingface.co]
"Stabe(sic) Diffusion", IE: Stable Diffusion: https://en.wikipedia.org/wiki/... [wikipedia.org]
DALL-E 3: https://openai.com/index/dall-... [openai.com]
Re: (Score:2)
NAK
Re: (Score:2)
I'm saying you're being daft on purpose. What part of my message wasn't properly received? Or are you just Naked At Keyboard? :-) You seem to do that a lot:
https://slashdot.org/comments.... [slashdot.org]
https://slashdot.org/comments.... [slashdot.org]
https://slashdot.org/comments.... [slashdot.org]
https://slashdot.org/comments.... [slashdot.org]
https://slashdot.org/comments.... [slashdot.org]
Re: (Score:2)
This seems to be a common occurence on /. lately. I got into the same sort of interaction with someone a few weeks back. He just wanted to argue with everything anyone said. And then you go look at his history and that seems to be all he could do. And each time it was the same format. At least the guy arguing with me didn't 'break' once he couldn't argue anymore. That guy would just keep going forever, even if you said the sky was blue, he was going to argue about it.
Re: (Score:2)
Certainly not limited to Slashdot, but a general effect of computers, me thinks. In the form of a joke, I'm sure that more than 35 years ago I was saying something like "Too much computer use is bad for mental health." But I didn't cut down, and perhaps I've become one of the poster children for the problem?
However I also think it's largely a result of declining literacy, so I should cite a book. Best I've recently read on this topic is The Anxious Generation by Jonathan Haidt. I'm planning to write a rev
Re: (Score:2)
This is one of the worst AI models I've seen.
Re: (Score:2)
Z^-1
Re: (Score:2)
c/Not much evidence of bibliophilia/Too much evidence of bibliophobia/
Re: (Score:2)
*sigh*
c/Deep Seek/DeepSeek/
c/DL/DS/
Actually I sort of wish the Preview had an AI-empowered grammar and spelling check... Maybe a sanity check, too?
Define Artificial Intelligence (Score:2)
AI? What's that? (Score:2)
Nothing more to see here folks, move along
Where's the "none" option? (Score:5, Insightful)
Or at least I could ask CowboyNeal ...
Re: (Score:2)
These days, it's hard to avoid using at least one or two models. Do you shop on Amazon? Then you're using Rufus, which automatically summarizes reviews, whether or not you actually read those reviews. Other shopping sites do similar things with AI models. If you use Google or Bing, you're using Gemini or GPT4, respectively, unless you went to the trouble of disabling AI search results.
It's only going to get *harder* to avoid generative AI, wherever you go on the web.
Re: Where's the "none" option? (Score:1)
Re: (Score:2)
Apparently, my luck with AI summaries is much better than yours. I find them to be a tremendous time saver. Yes, they do get it wrong sometimes, but in my experience, less often than Amazon reviews themselves are off the mark.
Re: (Score:2)
Apparently, my luck with AI summaries is much better than yours. I find them to be a tremendous time saver. Yes, they do get it wrong sometimes, but in my experience, less often than Amazon reviews themselves are off the mark.
I agree. Those summaries are actually what I view as the best use-case for LLM. The consequence of inaccuracy is low, the original material is right there and is easily spot-checked for obvious issues.
I'm pretty anti-AI as it exists today but this is not where it's falling down and untrustworthy.
Re: Where's the "none" option? (Score:1)
I donâ(TM)t think the pollâ(TM)s definition of âoeusingâ is the same as yours. Applying your logic, one might say everyone uses cocaine. ;)
Re: (Score:2)
I think you are referring to research that found that *all* of the money tested, contained trace amounts of cocaine.
This is not the same as how we all "use" AI in things like search results. This AI is more than a "trace" amount, it's front and center. Do people use that AI summary at the top of the search results? Do they read the AI summaries of the Amazon reviews? Probably, yes. That's a lot more consequential than a "trace" amount of cocaine that is not actually ingested and doesn't actually have an imp
Re: Where's the "none" option? (Score:1)
No, I wasnâ(TM)t quoting that, I meant literally.
I donâ(TM)t agree that I use a tool thatâ(TM)s being used by someone to provide any services to me. If we consider that most policemen in Brazil sniff coke, and they provide security to society, the latter uses coke. :D
Anyway, I read customer reviews, not these summaries. I am not even sure where they are.
And I donâ(TM)t use any LLM out there, I donâ(TM)t think they âoeadd valueâ to whatever I do in my life.
Re: (Score:2)
Well of course, if you don't use LLMs, you wouldn't think they add value. You've got to use them to find the value, or to know what kind of value they are capable of providing. And yes, they do provide significant time savings, if you learn how to use the tool.
Re: Where's the "none" option? (Score:1)
Iâ(TM)ve tried ChatGPT. For what I do, it sucks. Takes more time to explain what I need than doing it myself.
Re: (Score:2)
It does take a little trial and error to figure out how to leverage AI. It is, after all, a tool, and like all tools, you have to learn how to use it. I wasn't very good at AI prompts when I started either. But you do get better at it with practice.
Re: (Score:2)
Do you even read the crap you write? "â(TM)â(TM)â(TM)â(TM)" Time to use proper keyboard settings.
Re: (Score:1)
Do you even read the crap you write? "â(TM)â(TM)â(TM)â(TM)" Time to use proper keyboard settings.
Time for /. to support Unicode in the comments.
MOBI ! (Score:4, Informative)
Works every time (almost), costs me nothing except a little time, has a time-proven record (I'm over 60 years old now and still getting smarter every day, I think it's called wisdom), and does not hallucinate (except occasionally at parties).
Re: MOBI ! (Score:2)
I have done analysis on eye witness testimony in trials. Honestly, if those witnesses are representative of humans generally, the LLM hallucination rate is already well below the human rate. Many lawyers I have met have suggested maybe we should outright ban human testimony at trial because of the frequency of memory being unreliable.
Re: (Score:2)
Yep, same here. "None" it is. More effort than they are worth. May give them a chance again in a few years.
MFB (Score:3, Funny)
copilot (Score:3)
Because it's part of M365 and the company is already paying for us to have access. And they block all the others. So I only use the others when I'm on my personal computer or I drop the VPN on the work computer.
Re: copilot (Score:2)
CoPilot reportedly uses ChatGPT under the hood. So add your vote to that camp.
Obviously Missing (Score:5, Informative)
I don't
Re: (Score:2)
I don't
Beat me to it.
I used ChatGPT once, does that count? (Score:1)
Wetware only (Score:2)
I only run my AI-models on wetware. It still outperforms all other models ;)
Encyclopedia Britannica (Score:1)
granite-code (Score:2)
My employer doesn't want our code to escape to the cloud, so we can only use local models. The granite-code models from IBM run on ollama, and they're sometimes helpful, but it doesn't feel as magical as some people describe. Maybe I'm too old to be too excited. IDK. They work, and sometimes they offer some better, more complete autocomplete, but I feel like I'm missing something, because it just isn't that amazing.
Re: (Score:2)
I'm guessing AI functions as an active template maker for many folks. Can't just list out some bullet points and tell AI what the message intention is and off it goes.
If you have learned to think in hierarchal information format, I doubt AI would be all that useful in work.
Re: granite-code (Score:2)
All of them. (Score:3)
I installed every AI plugin there is in my IDE so they can all argue.
Re: (Score:2)
Just install one with a thinking/reflection state and crank up the settings and it'll argue with itself without even needing others
Other: Uh– (Score:2)
A Blob 'O Stardust I've been gifted with.
Seems to be working ok.
Cowboy Neal (Score:3)
Cowboy Neal is my AI.
Re: (Score:2)
Confirmed!
Re: (Score:2)
Where's the upvote? :)
Web search summary (Score:2)
The only AI I have (knowingly) used at all is the web search summary that comes up at the top of a Bing or Google search. I have no idea which AI model they use, nor do I care. I can say that given the specific technical nature of my searches (eg. SolidWorks, DraftSight, Bambu), they are usually less helpful than the actual search results including forums, wikis and help files/articles.(OEM or 3rd party).
Grok 3 (Score:2)
Recently released. This model is amazing, and has already helped me narrow down a neuropathy issue in my arm.
But mostly I use it for planning TTRPG sessions. The recommendations are good and it takes instructions quite well.
Re: (Score:2)
I wonder if I could fire up a few AIs and have a TTRPG campaign with them. Would be nice, the snacks are mine, getting together would be easy, and if I got killed off I'd delete the MFers.
Other (Score:1)
Qwen2.5 (Score:2)
Llama 3 is good, but I don't have enough VRAM to run a decent quant. Mistral Small 24B (the recent release) is really good for its size.
Moving away from OpenAI (Score:2)
I was ChatGPT subscriber as soon as I could get an account, and also have an API account. But in the last few weeks I've almost stopped using both.
Recently, I've just found Grok better than ChatGPT as a chatbot. I've also just discovered Cline for coding (I'm not a dev, but am interested in hobbyist tinkering) and am finding that Claude 3.5 Sonnet is considerably more reliable than OpenAI, at least in the Cline environment. Which leaves OpenAI out in the cold, for me.
Copilot (Score:1)
Nowadays I mostly use the copilot summaries in my web searches (I like it because unlike Google, they add the data origin with a few links, and I can click and read the whole article if I need clarification).
I've also used Copilot occasionally to write some SQL Queries with complicated nested joins I had trouble with, which I then modified adding more columns or parameters.
Why would I need at AI model? (Score:2)
Re: (Score:2)
I Don't Use AI (Score:2)
Re: I Don't Use AI (Score:2)
The same reason I use spellcheck, spreadsheets, the error squiggles in an ISE, or tab completion. This is just the next gen of those tools, and it is quite a leap.
Only what I can run locally. (Score:2)
Since I don't have an industrial scale server with half a terabyte of RAM, I'm not running DeepSeek-R1:671b, but I do run R1:70b and Llama 3.3. I only run models locally, whether LLMs or diffusion models. This is why I still haven't gotten into any sort of AI music, everything that isn't a complete toy exists only in the Cloud. Also I run FLUX.1 models fairly frequently (not every day, but every two or three days for several hours at a stretch) and this machine was specifically built to accomplish this on a
Eliza (Score:3)
* What AI models do you usually use most?
> We were discussing you, not me.
* But I need advice from an advisor as I can not decide myself which AI to use.
> Say, do you have any psychological problems?
* I guess I am... Indecisive?
> Do you believe it is normal to be indecisive?
* you tell me, you're the AI here!
> Does it please you to believe I am the AI here?
* argh
Mostly ChatGPT, but... (Score:2)
I have a paid (private) subscription to ChatGPT. However, I have local installations of Mixtral and DeepSeek, and I use Grok sometimes as well. It's interesting to see the variation in results. When my current (annual) subscription to ChatGPT runs out, I will re-evaluate. Things are moving so fast that likely some other AI will be a better choice.
The biggest problem I note: I use AI mostly in German, and lots of AIs are not very good outside of English.
ChatGPT (Score:3)
Actually India (Score:2)
None (Score:2)
They all suck.
SDXL and Flux AI (Score:2)
Pfft. (Score:2)
None of them, and it's going to stay that way.
Nothing (Score:1)
Nothing, they're rubbish
RE: What AI models do you usually use most? (Score:1)
No AI for me, I prefer RI (Score:2)
I prefer to use RI not AI
Are you aware that... (Score:2)
Don't know it it's Gemini (Score:2)
My best AI model (Score:2)
Idiots use AI. (Score:2)
Sorry, only idiots use AI to answer questions. Find your answer at least on wikipedia.