Vision + text · served with vLLM · ↓ 87.6K downloads a month · View on Hugging Face
RedHatAI/gemma-4-31B-it-NVFP4 needs about 37 GB of GPU memory for conversations of up to 8,192 tokens. It runs on 4x RTX 3060 from $0.34/h right now; best value: 2x RTX 5090 at $1.45/h.
Weights
23 GB
GPU memory needed
about 37 GB
Conversation length
8,192 tokens
Model allows
262,144 tokens
Memory · disk
19.6 GB · 66 GB
Architecture
Gemma4ForConditionalGeneration
Licence
apache-2.0
Versions of gemma-4 31B
The same model packaged differently: a quantized version needs less GPU memory and runs on a cheaper machine.
$1.45/h from your balance while it runs, 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 gemma-4-31B-it-NVFP4 need?
About 37 GB for conversations of up to 8,192 tokens: 23.27 GB of weights, the cache for one full-length conversation and the runtime's own memory.
What is the cheapest GPU to run gemma-4-31B-it-NVFP4?
Right now 4x RTX 3060 at $0.34 an hour; 2x RTX 5090 at $1.45 an hour is the best value for its speed. Prices are live offers from GPU clouds, checked every 15 minutes.
Can gemma-4-31B-it-NVFP4 run on an RTX 4090?
Not on one: it needs about 37 GB and an RTX 4090 has 24 GB. Use a bigger GPU, or a smaller (quantized) version of the model.
How do I deploy gemma-4-31B-it-NVFP4 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.
Prices are per hour for the whole machine and its disk, right now. Speed is Vast's DLPerf score of the machine, relative to the cheapest. Best value: the most speed per dollar. Only machines with a Vast reliability of 90% or more that download at 200 Mbit/s or more are offered: the machine fetches the model itself.