• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

Best LLMs/'AI's

Tams

Member
I couldn't find an existing thread, so please merge this if there is one.

What are people's experiences of using LLMs/'AI's?

I'm thinking of paying for one, or using one of those offers of credits for various ones.


My experience of the free ones have been:

  • ChatGPT: Overall very good, but very easily sycophantic.
  • Grok: Very good, and from what I've gotten from it, balanced.
  • Copilot: Useless.
  • Claude: Excellent for data analysis and able to produce graphs when the others failed. It was never able to produce consistent graphics though.
 
Last edited:
Pretty hit or miss, mostly miss. They are all incredibly sycophantic and are clearly designed to retain user session.

-ChatGPT: Great for research. Pretty much Google Research skip. Useful for medical but also dangerous.

- Gemini: Better at
analyzing scenarios and more creative.

- xAI: Too robotic. Good for technicals.

- Copilot: Pretty much GPT but worse in every category

-Lumo: About as good as GPT 5 mini.

The difference between models is basically 15%, they give you different awnsers for questions but you can see they are all trained on the same data and all have the same issues. They all fall apart for professional work, making them much more useful as - unironically, chatbots than actual tools for productvity.

Nowdays I use Lumo since none of these LLMs are private whatsoever.
 
Last edited:
ChatGPT has been great for some workflow things. Not quite as good for very niche research, but I'm generally familiar enough with AI to be able to spot hallucinations and verify sources at this point.
 
I use the various Grok personas in my Tesla, kids love the trivia mode. I can't imagine every paying for one though. I started using the Sora app recently but that's for videos, it's amazing.
 
I like Claude Sonnet 4.5 for code generation. OpenAI GPT-5 is pretty good for deep analysis. OpenAI GPT-o1 Pro is amazing for lots of stuff if you don't mind paying for the access.

I like to run gpt-oss locally. It fits on my 4070 Ti Super and my M4 MacBook Pro. It performs on par with GPT-4, which is actually great for local AI.
 
These days, I use Claude (Code) extensively (usually for scaffolding and sanity checking) for coding and writing Ansible playbooks and Molecule tests. It's a phenomenal tool that helps smooth over the busywork and serves as a fresh pair of eyes.
 
I use Amazon Q a lot, but it's just Claude's model w/ a wrapper CLI and IDE integration. It is better than using Co-Pilot for Developers in my experience.

But if you aren't a software dev you'll probably have never heard of Q.
 
Im notnreally a fan of Grok. Im sure it will get better but just doesn't feel anywhere near as good as chap got.

I havn't used direct comparisons asking them the same questions. I've just never walked away from satisfied. It's Image generation is terrible. Although it will now animate pictures for you.

For me gpt use it all the time.
 
I use:
ChatGPT 5 for daily things
Grok for images and short videos
Perplexity for correcting ChatGPT 5
HeyGen for virtual avatars
Elevenlabs for Speech
 
Only tried the ai search of google and it's crap, and the voice chat of gpt but it feels like a glorified talking search engine more than talking with a real person.

Sora seems to be the best for now but i never tried it.
 
How do Grok and Gemini compare for general, everyday use?

I currently have Gemini Pro, but I'm thinking of switching to Grok
 
We have been using ChatGPT and Copilot and the more I use them, the more I dislike them. I feel like they are getting things wrong more and more lately. I have gone back to regular ol' Google for IT troubleshooting. I have also found Gemini is way better for this as well.
 
I use Claude and ChatGPT 4o & 5 professionally and Grok in my personal life.

While LLMs are the new hotness, I still find ML a better solution for many of the problems I'm trying to solve in my job. It's also way cheaper at scale.

In my experience, Grok is by far the most enjoyable AI to converse with and produces excellent responses.
 
Last edited:
robert duvall GIF


Best AI is a dead AI
 
I use ChatGPT a lot, but Gemini is actually pretty good for troubleshooting Windows server issues. It help me figure out stuff that ChatGPT just couldn't. Claude isn't bad either, but I haven't used that in a while.
 
Well that's all of them.
If AI aren't alive then how did the albanian AI mothered 84 baby at 12 years old like every other albanian women ?

Article:
Albania's prime minister, Edi Rama, has announced that Diella, the world's first AI minister, is "pregnant with 83 children".


Jokes aside you are right, and people tend to forget too easily that ai isn't alive and doesn't think !
 
For coding or for any highly technical conversations (example: uploading the latest research paper in machine learning, full of formulas, and wanting to talk it out / discuss the reasoning and work through the decisions made), I always use Claude (paid; and my coding agent use is paid by my employer so I can go wild on that if desired).

For quick everyday conversations I use ChatGPT (paid). Its voice mode is also fantastic if you want to put the phone down while doing something, eg when I was working on our boat motor and wanted to ask live about some things to make sure I was thinking correctly about a problem.

For ultra speed (eg, if you want to script something like iterating over large datasets etc), the Gemini models are amazing. Gemini Flash 2.5 is the best "super fast but still useful" model.

Grok is the standout for current events. If you want to quickly learn why everyone is talking about something, or for instance get an update on a live case or situation, it does a fantastic job of pulling and synthesizing all the latest news and discussions.

So:
Claude: technical intellect
ChatGPT: general intellect
Gemini: speed
Grok: current events / pulse of the internet
Chinese models (eg. Qwen): best for open-source hackability if you want to custom tune a model for your purposes
 
Last edited:
Claude is amazing for creative writing also.

We use it all the time for copywriting at my job, its on par with some of our best staff guys. Its crazy.

It evolved so fast too, 12 months ago it was solid but I could still distinguish it from our human writers. Now its really really hard to tell.
 
Last edited:
I use ChatGPT for random day-to-day questions. I use Claude for coding. I use Grok as a fact-checker or context-giver on X (the integration with X is pretty amazing).

I don't really care about gemini and I wouldn't even put Microsoft co-pilot in the same ballpark as the aforementioned systems. They're trying some Copilot integration with Office 365 at the place I work and it typically is absolutely retarded stuff like "hey I noticed you're composing a mail to [person], maybe you wanna add this random excel file you've been working on yesterday as an attachment to this mail?". Just nonsense.

If you're feeling frisky you could install LM Studio or Ollama and download Mistral or Deepseek for some open source shenanigans.
 
I only use deepseek for an offline Ai for my home automation, it works wonderful and it doesn't have acces to internet.

I also use duck AI for sone small research tha always end up underwelming.
 
I've tried GPT, Grok, and Gemini. Grok is more expensive than GPT but performed worse, Gemini seemed alright, and GPT seemed to be all around the best. They all kinda suck to some degree though. There hasn't been a big improvement in a long while, at least as far as the language models go.
At the very least, GPT isn't that expensive for what you get, and it isn't really as sycophantic any more, though it does like to lecture you depending on what you are talking about.
 
Top Bottom