Which GPU runs which model
One page per model: how much VRAM it needs, which of our cards fit, and the hourly price you will actually be billed, read 2026-09-15.
Llama 3.1 8B8B
4.9 GB at Q4_K_M
from $0.21/h on RTX 3090 24G
Qwen3 8B8B
5 GB at Q4_K_M
from $0.21/h on RTX 3090 24G
Qwen3 14B14B
9 GB at Q4_K_M
from $0.21/h on RTX 3090 24G
Phi-4 14B14B
9 GB at Q4_K_M
from $0.21/h on RTX 3090 24G
Mistral Small 3 24B24B
14 GB at Q4_K_M
from $0.21/h on RTX 3090 24G
Gemma 3 27B27B
17 GB at Q4_K_M
from $0.21/h on RTX 3090 24G
Qwen3 32B32B
19 to 20 GB at Q4_K_M
from $0.59/h on RTX 5090 32G
DeepSeek-R1-Distill-Qwen-32B32B
19 to 20 GB at Q4_K_M
from $0.59/h on RTX 5090 32G
QwQ 32B32B
19 to 20 GB at Q4_K_M
from $0.59/h on RTX 5090 32G
Llama 3.3 70B70B
40 to 43 GB at Q4_K_M
from $1.09/h on A800 80G
Qwen2.5 72B72B
41 to 43 GB at Q4_K_M
from $1.09/h on A800 80G
DeepSeek-R1-Distill-Llama-70B70B
40 to 43 GB at Q4_K_M
from $1.09/h on A800 80G
gpt-oss-20b21B MoE (3.6B active)
13 GB at native MXFP4
from $0.21/h on RTX 3090 24G
gpt-oss-120b117B MoE (5.1B active)
63 GB at native MXFP4
from $1.09/h on A800 80G
FLUX.1 dev12B + T5-XXL encoder
12 to 16 GB at FP8
from $0.21/h on RTX 3090 24G
FLUX.1 schnell12B + T5-XXL encoder
12 to 16 GB at FP8
from $0.21/h on RTX 3090 24G
Stable Diffusion XL2.6B + encoders
7 GB at FP16
from $0.21/h on RTX 3090 24G
Whisper large-v31.5B
3 GB at FP16
from $0.21/h on RTX 3090 24G