GPUs would go so hard 🔥🔥🔥

Very much possible

https://x.com/justjake/status/1772005308052943005?s=46&t=00dq48d1c_GjD9m8fXEJ8w

However, GPUs have a liquidity problem (expensive for us to just keep em idle)

So, if you're interested, please upvote and once we hit enough demand, we can look into rolling this out broadly

We need this!

+1!

would be able to try this out

+1

+1

+1

+1

+1

+1

+1

+1

Yesss +1

+1

+1

there are news on this?

to use Ollama on Railway, it would be a perfect combo!

I would really need this

+1

+1

+1000

Hoping to move some GPU heavy services to railway ASAP.
+1

+1

+1 need GPUs

I currently use railway (just on hobby plan at the moment but definitely would upgrade), but am now building AI Image Generation applications with need for GPU. Would definitely prefer to stick with railway as, of all the deployment platforms I've tried, you guys are the most streamlined and easy to use.

Do we have any update on GPU services becoming available on railway? is there enough interest/resources to make it worth it yet for you to "flip that switch"? (I'm aware the implementation is significantly harder than that, you guys just make everything look so easy, thank you for that by the way!)

+1

I also would use this. understand the liquidity issues. I am looking to use high fidelity image upscaling. i wonder if there's a secure shared model that could allow us to not need per-user dedicated instances? This can allow us to scale as needed, without incuring massive idle costs for you.

Updates here?

+1

We're building an AI tool that uses a combination of local and public LLMs. CPU only workflow is very slow for some use cases so will need to purchase GPU capability in the next 2-4 months. Would greatly prefer to stay within Railway if possible

I would have use cases for GPUs as well. My scale would not occupy it too much though