Sure, llama.cpp is the real thing, ollama is a wrapper... I would never want to use something like ollama in a production setting. But if I want to quickly get someone less technical up to speed to develop an LLM-enabled system and run qwen or w/e locally, well then its pretty nice that they have a GUI and a .dmg to install.
Since the new multimodal engine, Ollama has moved off of llama.cpp as a wrapper. We do continue to use the GGML library, and ask hardware partners to help optimize it.
Ollama might look like a toy and what looks trivial to build. I can say, to keep its simplicity, we go through a deep amount of struggles to make it work with the experience we want.
Simplicity is often overlooked, but we want to build the world we want to see.
I knew a startup that deployed ollama on a customers premises and when I asked them why, they had absolutely no good reason. Likely they did it because it was easy. That's not the "easy to use" case you want to solve for.
We can obviously disagree with their priorities, their roadmap, the fact that the client isn't FOSS (I wish it was!), etc but no one can say that ollama doesn't work. It works. And like mchiang said above: its dead simple, on purpose.
(any differences are small enough that they either shouldn't cause the human much work or can very easily be delegated to AI)
Then you want to swap models on the fly. llama-swap you say? You now get to learn a new custom yaml based config file syntax that does basically nothing that the Ollama model file already does so that you can ultimately... have the same experience as Ollama but now you've lost hours just to get back to square one.
Then you need it to start and be ready with the system reboot? Great, now you get to write some systemd services, move stuff into system-level folders, create some groups and users and poof, there goes another hour of your time.
Why does this matter? For this specific release, we benchmarked against OpenAI’s reference implementation to make sure Ollama is on par. We also spent a significant amount of time getting harmony implemented the way intended.
I know vLLM also worked hard to implement against the reference and have shared their benchmarks publicly.
GGML library is llama.cpp. They are one and the same.
Ollama made sense when llama.cpp was hard to use. Ollama does not have value preposition anymore.
The models are implemented by Ollama https://github.com/ollama/ollama/tree/main/model/models
I can say as a fact, for the gpt-oss model, we also implemented our own MXFP4 kernel. Benchmarked against the reference implementations to make sure Ollama is on par. We implemented harmony and tested it. This should significantly impact tool calling capability.
Im not sure if im feeding here. We really love what we do, and I hope it shows in our product, in Ollama’s design and in our voice to our community.
You don’t have to like Ollama. That’s subjective to your taste. As a maintainer, I certainly hope to have you as a user one day. If we don’t meet your needs and you want to use an alternative project, that’s totally cool too. It’s the power of having a choice.
Is there a schedule for adding additional models to Turbo mode plan, in addition to gpt-oss 20/120b? I wanted to try your $20/month Turbo plan, but I would like to be able to experiment with a few other large models.
GGML is llama.cpp. It it developed by the same people as llama.cpp and powers everything llama.cpp does. You must know that. The fact that you are ignoring it very dishonest.
Nope…
Where can I learn more about this? llama.cpp is an inference application built using the ggml library. Does this mean, Ollama now has it's own code for what llama.cpp does?
We benchmarked vLLM and Ollama on both startup time and tokens per seconds. Ollama comes at the top. We hope to be able to publish these results soon.
If you can't get access to "real" datacenter GPUs for any reason and essentially do desktop, clientside deploys, it's your best bet.
It's not a common scenario, but a desktop with a 4090 or two is all you can get in some organizations.