Qwen

Qwen3.5 9B

Small 9B multimodal Qwen model with native 262,144-token context, widely used for self-hosting.image text to textapache-2.06.4M downloadsHugging Face

Cheapest GPU in stock that fits

A single GPU with at least 22 GB for inference.

NosanaIn stock · 9 freeStart right away

Nosana network, hosts worldwide

Details

Stock and price checked 16 min ago.

Self-service: you start it yourself, with no order or sales contact.

nvidia-3090

Full price

€0.17

per hour, whole instance, 1 GPU

From model cards and vendor documentation. Lower precision needs less memory, with slightly lower quality. Estimates are marked.

What you want to do	Precision	Minimum	Setup	Basis
Run the model vLLM recipe minimum for this repo's BF16 checkpoint.	BF16, full	22 GB	1x 24 GB GPU or Intel Arc Pro B60/B70 (per vLLM recipe)	Published
Run the model vLLM recipe minimum for separate repo RedHatAI/Qwen3.5-9B-FP8-dynamic.	FP8, 8-bit	11 GB	—	Published
LoRA fine-tuning Arithmetic from BF16 weight size (17.98 GiB = 9.65B parameters x 2 bytes): weights x 1.3, our standard estimate. Excludes activations beyond the overhead factor; batch size not assumed.	BF16, full	23.4 GB	—	Estimated
Full fine-tuning Arithmetic from BF16 weight size (17.98 GiB = 9.65B parameters x 2 bytes): weights x 4 states x 1.2, our standard estimate. Excludes activations beyond the overhead factor; batch size not assumed.	BF16, full	86.3 GB	—	Estimated