Workstation ×2 · FP8
$4.8/hr $0.0013/sec · $3,504/mo at 24/7
g4-standard-96 — two Blackwell GPUs.
VRAM pool
2× NVIDIA RTX PRO 6000 Blackwell 96GB
Compute die
NVIDIA RTX PRO 6000 Blackwell 96GB — FP8 tensor cores.
GDDR7 memory
192 GB of GDDR7 across the 2 GPUs.
PCIe Gen5 interconnect
PCIe Gen5 host links — 2 GPUs, each on its own link.
Per-second metering
$4.8/hr → 0.0013/sec. Pause the deployment and the meter hits $0.
Scale-to-zero
Autoscale + region pins (GLOBAL / US / EU / APAC) on the control plane.
Memory budget
192 GB of VRAM on this preset. 13-30B dense at FP8 plus LoRA fine-tunes and several concurrent workers.
Typical loads
production chat farms, nightly fine-tunes, concurrent batch jobs. Fine on: torch, JAX, raw CUDA, axolotl, vLLM, sglang, ComfyUI.
Billing math
$4.8/hr is $0.0013/sec while running and $0 while parked; park/wake freezes and restores state instead of paying for idle.
Best fit
192 GB across 2 RTX PRO 6000 Blackwell cards at $4.80/hr, $2.40 per GPU-hour — two separate 96 GB GDDR7 budgets on two PCIe Gen5 links. At FP8 that is two 70B checkpoints side by side (~70 GB of weights each, ~26 GB left per card) or one ~140B model split across both cards (rule of thumb).
Trade-off
g4-4x is $9.60/hr: two more cards and 384 GB total, still $2.40/hr per GPU. The step up adds 96 GB budgets and PCIe Gen5 links; per-card VRAM and the per-GPU price do not change, so it is pure capacity scaling.
$4.8/hr on runaii.cloud, metered per second: $0.0013/sec while the VM runs and $0 while parked. A month of 24/7 uptime would be about $3,504, and park/wake means busy-fraction pricing in practice.
Yes. Per-second metering with park/wake: idle freezes the VM state into a signed artifact and stops the meter the same second; the next request wakes it byte-exact
192 GB of VRAM. 13-30B dense at FP8 plus LoRA fine-tunes and several concurrent workers.
In total, yes: 192 GB against 96 GB, at $4.80/hr against $2.40/hr. Per card it is still 96 GB of GDDR7, because each card is a separate PCIe Gen5 device.
$2.40/hr per GPU — identical to g4-1x, so the 2× row is linear pricing with no multi-GPU discount.
Two independent 70B FP8 checkpoints (~70 GB of weights each) or one ~140B checkpoint split across both cards, each case with ~26 GB of headroom per card (rule of thumb).
RTX PRO 6000 slices
from $0.30/hr · 12-48GB
Blackwell slices for light jobs
1× NVIDIA T4 16GB
$0.44/hr · FP16
Starter
1× NVIDIA L4 24GB
$0.87/hr · FP8
Efficient
2× NVIDIA L4 24GB
$1.8/hr · FP8
Efficient ×2
4× NVIDIA L4 24GB
$3.6/hr · FP8
Efficient ×4
8× NVIDIA L4 24GB
$7.2/hr · FP8
Efficient ×8
1× NVIDIA RTX PRO 6000 Blackwell 96GB
$2.4/hr · FP8
Workstation
4× NVIDIA RTX PRO 6000 Blackwell 96GB
$9.6/hr · FP8
Workstation ×4
8× NVIDIA RTX PRO 6000 Blackwell 96GB
$19.2/hr · FP8
Workstation ×8
8× NVIDIA B300 288GB
$96/hr · FP8
Full Precision
8× NVIDIA B200 180GB
$80/hr · NVFP4
Throughput
4× NVIDIA B300 288GB
$48/hr · NVFP4
Minimal
4× NVIDIA B200 180GB
$40/hr · NVFP4
Efficient
16× NVIDIA GB300
$184/hr · FP8
Frontier Max