Efficient · FP8
$0.87/hr $0.0002/sec · $635.1/mo at 24/7
Best price/perf up to ~13B + video.
VRAM pool
1× NVIDIA L4 24GB
Compute die
NVIDIA L4 24GB — FP8 tensor cores.
GDDR6 memory
24 GB of GDDR6 across the GPU.
PCIe Gen4 interconnect
PCIe Gen4 host links.
Per-second metering
$0.87/hr → 0.0002/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
24 GB of VRAM on this preset. The catalog's price/perf sweet spot: ~13B dense plus video pipelines at FP8.
Typical loads
~13B chat plus video pipelines, embedding churn, RAG reindex. Fine on: torch, JAX, raw CUDA, axolotl, vLLM, sglang, ComfyUI.
Billing math
$0.87/hr is $0.0002/sec while running and $0 while parked; park/wake freezes and restores state instead of paying for idle.
Best fit
24 GB of GDDR6 on one PCIe Gen4 card at the catalog's FP8 precision, $0.87/hr. At FP8 weights cost about one byte per parameter, so a 13B checkpoint leaves roughly 11 GB of the 24 GB card for KV cache and activations (rule of thumb). One card means one worker: scale out to a second L4 only when this budget or this worker is the limit.
Trade-off
The next rung is l4-2x at $1.80/hr: a second 24 GB card and a second PCIe Gen4 link, 48 GB in total. Per GPU that is $0.90/hr against $0.87/hr here, so the second card buys capacity, not a discount. Stay on one L4 while a single 24 GB budget and one worker cover the load.
$0.87/hr on runaii.cloud, metered per second: $0.0002/sec while the VM runs and $0 while parked. A month of 24/7 uptime would be about $635, 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
24 GB of VRAM. The catalog's price/perf sweet spot: ~13B dense plus video pipelines at FP8.
About 13B parameters of weights at one byte per parameter (rule of thumb), leaving roughly 11 GB of the 24 GB card for KV cache and activations.
Yes — 1× NVIDIA L4 24GB on PCIe Gen4 at $0.87/hr. Multi-card L4 shapes start at l4-2x: 48 GB for $1.80/hr.
$0.87/hr — the same as its hourly rate, because the preset is one GPU. The 2× L4 row is $0.90/hr per GPU, $0.03/GPU/hr more than this card.
RTX PRO 6000 slices
from $0.30/hr · 12-48GB
Blackwell slices for light jobs
1× NVIDIA T4 16GB
$0.44/hr · FP16
Starter
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
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