RunpodReviews of runpod.io

half easy half hard to manage, but great great AI bot assitant

**RunPod Needs a GPU Reservation System — Not Just a Waiting List** I can understand RunPod's business model. Of course, the company wants its GPUs to be running 24/7 rather than sitting idle. But honestly, I would not recommend RunPod for serious projects if you have to wait for a GPU to become available. Switching to another GPU instance is not always a simple solution. You risk losing your data, configuration, installed dependencies, and hours of work. Data loss may not happen every time, but migration can be complicated and risky, especially when you are working on serious projects. There is a very simple solution that could benefit both RunPod and its customers: **a GPU reservation system.** Imagine a reservation calendar displayed next to the available GPUs. Users could see which time slots are occupied and which are available, then reserve a GPU for a specific period. The payment could be charged upfront, or automatically two hours before the reservation begins. Once the reservation is confirmed, the customer knows that the GPU will be available at the agreed time. This would give customers predictability, eliminate unnecessary waiting and stress, and make it easier to plan serious development work. At the same time, RunPod would have better visibility into future demand and could monetize its GPU capacity more efficiently. It is a straightforward idea. Instead of forcing customers to constantly check availability or migrate their projects between instances, why not let them book the computing resources they need? **RunPod, this is a win-win opportunity.** You could improve the customer experience, build customer loyalty, and make better use of your infrastructure. Perhaps someone at RunPod will hear this suggestion and turn it into a real feature.

it is totally functional and usable. sometimes i feel it is costly compare with other available sites. if the GPU and storage cost could be slightly reduced, then it will grab more audiance...cheers!

Overall it works as I'd expect, only issue is GPU availability, but I also get that it would be more expensive if they had a ton of idle GPUs just sitting there all the time.

Fairly easy to use (with the help of ChatGPT!) and quite intuitive. The servers are slow when performing operations, though downloading models is really fast. I’ve only worked with the Jupyter terminal, and it’s excellent. It would be better if some of the faster models cost a bit less: $2.07 an hour is a significant amount if you need to work on something for a long time. You have to migrate too often—even though the process is seamless since the server handles everything, you sometimes have to wait quite a while.

Constantly have to move my container as GPUs are not available. Need to get the Global Storage truly global, otherwise I have a bunch of dead pods, and Network storage I can't use.

GPUs are so expensive these days, right? Your service has been a total lifesaver for running popular generative AI. Thanks!

best in their offerings, i recently moved from a HORRIBLE GPU provider (salad) and run pod is costlier but boy are they worth it, extremely detailed, i get to see all the logs of everything which enabled me to fix stuff, pricing is on the higher side but i would want nothing more than stable service and honest people to have my money