Qubrid AI is an open-source AI platform spanning compute, inference, fine-tuning and retrieval. Serverless APIs run models without any infrastructure to manage; GPU virtual machines give dedicated endpoints with SSH access, configurable storage and auto-stop; and AI Factory covers bare metal and appliance scale-out. Hardware runs to NVIDIA H200 at 141GB, B200 at 180GB and H100 at 80GB, and the model catalogue includes MiniMax-M3, GLM-5.1 and 5.2 and the Qwen 3.6 and 3.7 series.
Two details are better than the category standard. Auto-stop on GPU VMs is a small feature that prevents the most common and most expensive mistake in rented compute, which is leaving an H200 running over a weekend – the bill from that dwarfs any hourly rate difference between providers. And an open-source platform layer matters for the same reason it matters in a knowledge base: the workloads you build here should be portable, and code you can read is what makes that true.
The model catalogue is worth reading closely, because it is weighted toward Chinese open-weight families – MiniMax, GLM and Qwen – which are strong models and also a procurement question at some organisations, so check your own policy before planning around them. Pricing is the bigger gap: the only figure on the page is $1 of free API credit on a $5 deposit, and per-token and per-GPU-hour rates sit behind a separate page. On a commodity where price is most of the decision, that should be the first thing you check rather than the last.
