87 comments

[ 2.1 ms ] story [ 79.2 ms ] thread
I got tired of repeating the same points and having to dig up sources every time, so here's the timeline (as I know it) in one place with sources.
> Ollama eventually added ollama run hf.co/{repo}:{quant} to pull directly from Hugging Face, which partially addresses the availability problem.

uh actually, _we_ did (generates a Docker-style manifest on the fly)

Hah! I did not know that, I'll update the article!
Thank you; it's an educating read for me, as someone who doesn't dwell in this space, but cares about FOSS in its true spirit.
Do they still not let you change the default model folder? You had to go through this whole song and dance to manually register a model via a pointless dockerfile wannabe that then seemed to copy the original model into their hash storage (again, unable to change where that storage lived).

At the time I dropped it for LMStudio, which to be fair was not fully open source either, but at least exposed the model folder and integrated with HF rather than a proprietary model garden for no good reason.

On a practical note if fumbles connection handling as to be unusable to download anything.
I think the biggest advantage for me with ollama is the ability to "hotswap" models with different utility instead of restarting the server with different models combined with the simple "ollama pull model". In other words, it has been quite convenient.

Due to this post I had to search a bit and it seems that llama.cpp recently got router support[1], so I need to have a look at this.

My main use for this is a discord bot where I have different models for different features like replying to messages with images/video or pure text, and non reply generation of sentiment and image descriptions. These all perform best with different models and it has been very convenient for the server to just swap in and out models on request.

[1] https://huggingface.co/blog/ggml-org/model-management-in-lla...

Llama-server which is part of llamacpp does this for a few months now
i had no idea about all this. especially the performance and bugs. thanks for informing me!
I prefer Ollama over the suggested alternatives.

I will switch once we have good user experience on simple features.

A new model is released on HF or the Ollama registry? One `ollama pull` and it's available. It's underwhelming? `ollama rm`.

No mention of the fact that Ollama is about 1000x easier to use. Llama.cpp is a great project, but it's also one of the least user friendly pieces of software I've used. I don't think anyone in the project cares about normal users.

I started with Ollama, and it was great. But I moved to llama.cpp to have more up-to-date fixes. I still use Ollama to pull and list my models because it's so easy. I then built my own set of scripts to populate a separate cache directory of hardlinks so llama-swap can load the gguf's into llama.cpp.

Koboldcpp is a single executable with a GUI launcher and a built in webui. It also supports tts, stt, image gen, embeddings, music creation, and a bunch of other stuff out of the box, and can download and browse HF models from within the GUI. That's pretty easy to use.
Least friendly you’ve used makes me think you’ve been spoiled. :)

Agreed ollama is a good intro but once you move beyond it starts to be a pain.

> No mention of the fact that Ollama is about 1000x easier to use.

The point of the article is not to expound on how user-friendly "Ollama" is. It's about exposing the deception and shameful moral low ground they took.

It feels like a bit of history is missing... If ollama was founded 3 years before llama.cpp was released, what engine did they use then? When did they transition?
The performance issues are crazy. Thanks for sharing this
I noticed the performance issues too. I started using Jan recently and tried running the same model via llama.cpp vs local ollama, and the llama.cpp one was noticeably faster.
Not sure why VLC doesn't do that.

It's a joke... but also not really? I mean VLC is "just" an interface to play videos. Videos are content files one "interact" with, mostly play/pause and few other functions like seeking. Because there are different video formats VLC relies on codecs to decode the videos, so basically delegating the "hard" part to codecs.

Now... what's the difference here? A model is a codec, the interactions are sending text/image/etc to it, output is text/image/etc out. It's not even radically bigger in size as videos can be huge, like models.

I'm confused as why this isn't a solved problem, especially (and yes I'm being a big sarcastic here, can't help myself) in a time where "AI" supposedly made all smart wise developers who rely on it 10x or even 1000x more productive.

Weird.

For most users that wanted to run LLM locally, ollama solved the UX problem.

One command, and you are running the models even with the rocm drivers without knowing.

If llama provides such UX, they failed terrible at communicating that. Starting with the name. Llama.cpp: that's a cpp library! Ollama is the wrapper. That's the mental model. I don't want to build my own program! I just want to have fun :-P

"LM Studio… Jan… Msty… koboldcpp…"

Plenty of alternatives listed. Can anyone with experience suggest the likely successor to Ollama? I have a Mac Mini but don't mind a C/L tool.

I think, as was pointed out, Ollama won because of how easy it is to set up, pull down new models. I would expect similar for a replacement.

> Ollama is a Y Combinator-backed (W21) startup, founded by engineers who previously built a Docker GUI that was acquired by Docker Inc. The playbook is familiar: wrap an existing open-source project in a user-friendly interface, build a user base, raise money, then figure out monetization.

    The progression follows the pattern cleanly:

    1. Launch on open source, build on llama.cpp, gain community trust
    2. Minimize attribution, make the product look self-sufficient to investors
    3. Create lock-in, proprietary model registry format, hashed filenames that don’t work with other tools
    4. Launch closed-source components, the GUI app
    5. Add cloud services, the monetization vector
The missing attribution pattern is nasty.
ollama is pretty intuitive to use still - dont see why will stop.
Have you ever tried going to the model registry and seeing that the model was recently updated? What updated? What changed? Should I re-download this 20GB file?

I guess if you're not frustrated with things like this then sure, no need to stop using it.

LM Studio is equally as simple, has all the same features, and none of the performance or lock-in problems of ollama.

If you only needed a single reason, how about kneecapping your performance by choosing ollama?

I see no mention of vLLM in the article.
Another scummy YCombinator project, one of many lately. Looks like no-one is left at the wheel, at least as long as the valuations (and hence money) keep coming in.
Alas people want convenience and don’t care about this sort of stuff.
Hmm..

  pacman -Ss ollama | wc -l                                                                                                              
  16
  pacman -Ss llama.cpp | wc -l
  0
  pacman -Ss lmstudio | wc -l
  0
Maybe some day.
Then again...

  zypper --no-refresh search llamacpp | tail -n5 | wc -l
  5
Sometimes Arch has the software you want at the version you want, other times it doesn't but other distros do. That's why there's half a billion distros instead of just one.
I was pretty big on ollama, it seemed like a great default solution. I had alpha that it was a trash organization but I didn't listen because I just liked having a reliable inference backend that didn't require me to install torch. I switched to llama.cpp for everything maybe 6 months ago because of how fucking frustrating every one of my interactions with ollama (the organization) were. I wanna publicly apologize to everyone who's concerns I brushed off. Ollama is a vampire on the culture and their demise cannot come soon enough.

FWIW llama.cpp does almost everything ollama does better than ollama with the exception of model management, but like, be real, you can just ask it to write an API of your preferred shape and qwen will handle it without issue.

Oh I was completely wrong about the model management stuff btw, llama-server has fully fledged model management baked in now, you just have to make an *.ini with your model configs (most models can do this themselves, I pointed qwen3.6 at the relevant part of the docs and it wrote me an ini with all of my model configs in about 2 minutes) and you can swap between models via api or a dropdown menu in the UI.