Workstation ×4 · FP8
$9.6/hr $0.0027/sec · $7,008/mo at 24/7
g4-standard-192 — four GPUs.
VRAM pool
4× NVIDIA RTX PRO 6000 Blackwell 96GB
Compute die
NVIDIA RTX PRO 6000 Blackwell 96GB — FP8 tensor cores.
GDDR7 memory
384 GB of GDDR7 across the 4 GPUs.
PCIe Gen5 interconnect
PCIe Gen5 host links — 4 GPUs, each on its own link.
Per-second metering
$9.6/hr → 0.0027/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
384 GB of VRAM on this preset. 70B+ dense checkpoints fit with room for KV cache (rule of thumb: about one byte per parameter at FP8).
Typical loads
multi-tenant inference, parallel eval harnesses, 70B+ dense serving. Fine on: torch, JAX, raw CUDA, axolotl, vLLM, sglang, ComfyUI.
Billing math
$9.6/hr is $0.0027/sec while running and $0 while parked; park/wake freezes and restores state instead of paying for idle.
Best fit
384 GB across 4 RTX PRO 6000 Blackwell cards at $9.60/hr, $2.40 per GPU-hour — four separate 96 GB GDDR7 budgets on PCIe Gen5. At FP8 that is four 70B checkpoints, or one ~280B model split four ways with ~104 GB left for KV cache (rule of thumb).
Trade-off
g4-8x is $19.20/hr: eight cards and 768 GB total, $2.40 per GPU-hour — double the cards and double the VRAM for double the hourly. Nothing changes per card, and there is still no NVLink between RTX PRO 6000 cards; each one is a PCIe Gen5 device.
$9.6/hr on runaii.cloud, metered per second: $0.0027/sec while the VM runs and $0 while parked. A month of 24/7 uptime would be about $7,008, 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
384 GB of VRAM. 70B+ dense checkpoints fit with room for KV cache (rule of thumb: about one byte per parameter at FP8).
384 GB — four 96 GB RTX PRO 6000 Blackwell cards, $9.60/hr, $2.40 per GPU-hour.
No — $2.40/hr per GPU on both rows. The 4× preset is exactly 4× the hourly of the single card, so you pay for capacity, not a discount.
At FP8 and one byte per parameter, a ~280B checkpoint split across four cards leaves ~104 GB of the 384 GB pool for KV cache, or four 70B checkpoints fit side by side (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
2× NVIDIA RTX PRO 6000 Blackwell 96GB
$4.8/hr · FP8
Workstation ×2
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