Efficient ×8 · FP8
$7.2/hr $0.0020/sec · $5,256/mo at 24/7
g2-standard-96 — full-node L4. g2-standard-96 — full-node L4.
VRAM pool
8× NVIDIA L4 24GB
Position
Priced per GPU-hour — see the launch wizard for live rates
Compute die
NVIDIA L4 24GB — FP8 tensor cores, tuned for g2-standard-96 — full-node l4.
HBM stack
8× NVIDIA L4 24GB of high-bandwidth attached memory — the whole model fits, no offload.
Interposer / NVLink
NVLink-C2C fabric so the 8 GPUs act like one pool.
Per-second metering
$7.2/hr → 0.0020/sec. Pause the deployment and the meter hits $0.
Scale-to-zero
Autoscale + region pins (GLOBAL / US / EU / APAC) on the control plane.
1× NVIDIA T4 16GB
$0.44/hr · FP16
Cheapest CUDA — experiments, small models
1× NVIDIA L4 24GB
$0.87/hr · FP8
Best price/perf up to ~13B + video
2× NVIDIA L4 24GB
$1.8/hr · FP8
g2-standard-24 — two-worker inference
4× NVIDIA L4 24GB
$3.6/hr · FP8
g2-standard-48 — multi-worker farm on one VM
1× NVIDIA RTX PRO 6000 Blackwell (1/8 GPU · 12GB)
$0.3/hr · FP8
g4-standard-6 — Blackwell slice for light jobs
1× NVIDIA RTX PRO 6000 Blackwell (1/4 GPU · 24GB)
$0.6/hr · FP8
g4-standard-12 — quarter GPU
1× NVIDIA RTX PRO 6000 Blackwell (1/2 GPU · 48GB)
$1.2/hr · FP8
g4-standard-24 — half GPU
1× NVIDIA RTX PRO 6000 Blackwell 96GB
$2.4/hr · FP8
g4-standard-48 — try any app from $2.40/hr (park/wake receipted D27)
2× NVIDIA RTX PRO 6000 Blackwell 96GB
$4.8/hr · FP8
g4-standard-96 — two Blackwell GPUs
4× NVIDIA RTX PRO 6000 Blackwell 96GB
$9.6/hr · FP8
g4-standard-192 — four GPUs
8× NVIDIA RTX PRO 6000 Blackwell 96GB
$19.2/hr · FP8
g4-standard-384 — full-node Blackwell
8× NVIDIA B300 288GB
$96/hr · FP8
Max quality for frontier dense models
8× NVIDIA B200 180GB
$80/hr · NVFP4
Best tokens/$ for large MoE models
4× NVIDIA B300 288GB
$48/hr · NVFP4
Smallest frontier-capable footprint
4× NVIDIA B200 180GB
$40/hr · NVFP4
Great for ≤70B dense models
16× NVIDIA GB300
$184/hr · FP8
For 1T+ parameter models