Exposing my own API from a rented GPU
Can I expose my own public API from a rented GPU?
Can I expose my own public API from a rented GPU?
Technically a dedicated machine can run your own serving stack, but public ingress is a separate networking question.
The current site promises SSH access and self-managed serving, but I do not see enough public detail to promise arbitrary inbound ports or a public IP setup.
So the safe FAQ answer is: yes for running your own API software, but confirm the networking/port requirements with AxForge if it must be reachable directly from the public internet.
Can the API generate music?
Yes. MiniMax Music 3 is currently listed on the serverless API.
The public price is €0.09 per track.
Is 4x RTX 5090 basically the same thing as a GB10 because both add up to 128 GB?
No. Four 32 GB cards do not automatically behave like one 128 GB GPU.
With four 5090s, software normally has to shard a model or workload across the cards. Each card still has its own 32 GB memory.
The GB10 has one shared 128 GB memory pool, which is simpler for models that need more than 32 GB in one addressable pool. Four 5090s can still be far more attractive when the workload scales well across multiple GPUs.
Sign in to reply.
Posting guidelines
Accounts that break these can lose forum access — paid plans included. Read the full guidelines →
Related topics