Rent an NVIDIA RTX PRO 6000 Max-Q GPU with 96 GB VRAM
Rent RTX PRO 6000 Max-Q cloud GPU servers by the hour on Cloudj: 96 GB VRAM, NVIDIA Blackwell and about 119 TFLOPS FP32 at a 300 W power limit. It packs flagship Blackwell memory into an efficient card, for AI teams who need 96 GB for 70B-class models, long contexts and large generative pipelines. Use CUDA 12.8 or newer builds of your software, pick a live configuration and connect over SSH with your key.
RTX PRO 6000 Max-Q specifications
VRAM per GPU
96 GB
Architecture
NVIDIA Blackwell
Compute capability
12
FP32 compute
≈ 119 TFLOPS
Memory bandwidth
≈ 1404 GB/s
Power limit
300 W
Values reported by the infrastructure provider for machines currently listed; real performance varies by machine.
Live RTX PRO 6000 Max-Q configurations and prices
From $1.04/h
incl. 32 GB disk · Live price, updated every 15 minutes
What you can run on the RTX PRO 6000 Max-Q cloud GPU
70B models at 8-bit
96 GB holds a 70B model at 8-bit, or a 4-bit 70B model with a very large KV cache.
Long-context and multi-user serving
1404 GB/s of bandwidth and 96 GB let vLLM or SGLang handle long prompts and many sessions.
Single-GPU 70B fine-tuning
Run QLoRA on 70B-class models with longer sequences than 48–80 GB cards allow.
Video and 3D generation
Load large video, 3D and image models together in ComfyUI without offloading.
Is the RTX PRO 6000 Max-Q the right GPU for your workload?
Choose the RTX PRO 6000 Max-Q when your workloads require 96 GB of VRAM and the Blackwell feature set on a single workstation card. Opt for a smaller consumer GPU if your models require less memory, or select a datacenter GPU if your workflow requires higher memory bandwidth than about 1404 GB/s.
RTX PRO 6000 Max-Q vs similar cloud GPUs
Guides for RTX PRO 6000 Max-Q
RTX PRO 6000 Max-Q cloud GPU: frequently asked questions
Rent the RTX PRO 6000 Max-Q cloud GPU by the hour
Pick a live configuration, choose a template and connect over SSH with your key. Billing stops when you destroy the instance.