@f As a rule of thump, I agree. But in the case of someone who can not or does not want to run things on self-owned hardware:

**What is the next best solution when it comes to running LLM queries?** Renting a GPU is hardly practical for most users.

I should have been more clear. The project I am working on clearly advocates to run the LLM inference locally. It would be nice to offer an alternative.

@n_dimension Yes, running the model directly is the desired use case for the selfhosted assistant I am creating (https://gitlab.com/RmMsr/relagent). And that works great.

To let people try it or if someone has no other option, I want to suggest something more suitable than just chatgpt, gemini, openrouter or whatever turns up on the internet that day. ;-)

@Alain Danke dir.

Ja, openrouter ist eine Option. Allerdings eher nichtssagende terms of service zu Datenschutz. Plus Verzicht auf Rechte bei opt-in zu Logging.

Parasail hab ich probiert. Keine sichtbare Kostenkontrolle. Und kein tool calling für gpt-oss. Olmo3 verhällt sich eigenartig unflexibel.

Pindown werd ich mal ansehen.

@duncan_bayne ovh.com (France) has been working well. Messy and overloaded user interface. But in terms of hosting they have a lot to offer.
I heared good things about gandi.net (also french) if you are not hunting for lowest prices.

All 11 australian providers are listed by IANA, but I have no comment on those: https://www.icann.org/en/contracted-parties/accredited-registrars/list-of-accredited-registrars?page=1&country=Australia

@taschenorakel I got this using gpt-oss-20b-mxfp4-GGUF:

```
#include <print>

int main() {
std::println("Hello, World!");
}
```

I am out of the loop with C++ for some time. So I am not sure on including <print> and the std namespace.

@iamlayer8 True. The model itself is stateless. The state (or context if you want) is all that the services memorizes and adds that to all requests. And that part is making those services more and more valuable.

Which goes in both ways. Usefulness for the user and better data for the provider. I agree privacy and information ownership is a major concern here.

Reading this blog post I learned how little innovation beyond the LLM itself was probably needed to get here: https://manthanguptaa.in/posts/chatgpt_memory/