Brendan McKeag

With MIG, we can partition RTX 6000 Pro cards into isolated 24 GB instances. Here's when it makes sense for your workloads.

Learn how to set up a real-world agentic system with our new Flash framework.

Flash is now generally available (GA) as a production-ready tool for running serverless GPU and CPU workloads in pure Python without needing Docker.

DeepSeek V4 is not the "Sputnik moment" R1 was, but it is the cheapest credible alternative to Claude Opus and GPT-5.5 that anyone has shipped thus far.

Runpod continues to add to its fleet and add new data centers to bolster supply offerings.

Runpod's platform continues to evolve, and now, so does your ability to keep tabs on where your GPU dollars are going. We've rolled out cost centers, a billing feature that lets you tag your Runpod resources with labels and track spend by team, project, or department. If you've ever gotten an invoice and thought "wait, who spun up all these pods?" this one's for you.

Our esteemed Discord community helper notrius built a single container that bundles dataset prep, model management, three training backends, two inference UIs, and a full control plane, so you can stop fighting dependencies and start creating. No more spinning up one pod for Comfy and another for training and kicking files back and forth through the CLI; you can run everything in a single pod now, using a single GPU,