How much GPU memory does MAI-DS-R1 need?

About 1509.3 GB for conversations of up to 8,192 tokens: 1342.06 GB of weights, the cache for one full-length conversation and the runtime's own memory.

What is the cheapest GPU to run MAI-DS-R1?

No machine on offer right now has enough GPU memory.

Can MAI-DS-R1 run on an RTX 4090?

Not on one: it needs about 1509.3 GB and an RTX 4090 has 24 GB. Use a bigger GPU, or a smaller (quantized) version of the model.

How do I deploy MAI-DS-R1 as an API?

Pick a machine on its GridPup page and press Deploy: GridPup rents the GPU, downloads the model from Hugging Face and serves it at an https address with keys and OpenAI's chat API, billed by the hour.

Models / MAI-DS-R1

microsoft/MAI-DS-R1

Text generation · served with vLLM · ↓ 643 downloads a month · View on Hugging Face

microsoft/MAI-DS-R1 needs about 1509.3 GB of GPU memory for conversations of up to 8,192 tokens. No machine on offer right now is big enough for it.

Weights: 1342 GB
GPU memory needed: about 1509.3 GB
Conversation length: 8,192 tokens
Model allows: 163,840 tokens
Memory · disk: 679 GB · 1517 GB
Architecture: DeepseekV3ForCausalLM
Licence: mit

Hardware and deploy

No machine available right now can serve the model: the largest its 8 GPUs have 179 GB each and the model needs about 190.6 GB on each.

Endpoint name

billed by the hour; stop it from its page. GridPup rents the machine, which downloads the model from Hugging Face (usually ready in 10 to 15 minutes), and gives you an https address, keys and a Playground.

Questions

How much GPU memory does MAI-DS-R1 need?: About 1509.3 GB for conversations of up to 8,192 tokens: 1342.06 GB of weights, the cache for one full-length conversation and the runtime's own memory.
What is the cheapest GPU to run MAI-DS-R1?: No machine on offer right now has enough GPU memory.
Can MAI-DS-R1 run on an RTX 4090?: Not on one: it needs about 1509.3 GB and an RTX 4090 has 24 GB. Use a bigger GPU, or a smaller (quantized) version of the model.
How do I deploy MAI-DS-R1 as an API?: Pick a machine on its GridPup page and press Deploy: GridPup rents the GPU, downloads the model from Hugging Face and serves it at an https address with keys and OpenAI's chat API, billed by the hour.

Loading…