Vision + text · served with vLLM · ↓ 58.2K downloads a month · View on Hugging Face
Qwen/Qwen3-VL-2B-Instruct-FP8 needs about 7.1 GB of GPU memory for conversations of up to 8,192 tokens. It runs on one RTX 3060 from $0.09/h right now; best value: RTX 5070 Ti at $0.22/h.
Weights
3.5 GB
GPU memory needed
about 7.1 GB
Conversation length
8,192 tokens
Model allows
262,144 tokens
Memory · disk
16 GB · 44 GB
Architecture
Qwen3VLForConditionalGeneration
Licence
apache-2.0
Versions of Qwen3-VL 2B
The same model packaged differently: a quantized version needs less GPU memory and runs on a cheaper machine.
Name and access:Qwen3-VL-2B-Instruct-FP8 · with a key
$0.22/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 Qwen3-VL-2B-Instruct-FP8 need?
About 7.1 GB for conversations of up to 8,192 tokens: 3.47 GB of weights, the cache for one full-length conversation and the runtime's own memory.
What is the cheapest GPU to run Qwen3-VL-2B-Instruct-FP8?
Right now one RTX 3060 at $0.09 an hour; RTX 5070 Ti at $0.22 an hour is the best value for its speed. Prices are live offers from GPU clouds, checked every 15 minutes.
Can Qwen3-VL-2B-Instruct-FP8 run on an RTX 4090?
Yes: it needs about 7.1 GB and an RTX 4090 has 24 GB.
How do I deploy Qwen3-VL-2B-Instruct-FP8 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.