Frontier Max · FP8
$184/hr $0.0511/sec · $134,320/mo at 24/7
For 1T+ parameter models. 1T+ parameter models.
VRAM pool
288GB ×16 = 4.6TB
Position
The frontier ceiling
Compute die
NVIDIA GB300 — FP8 tensor cores, tuned for 1t+ parameter models.
HBM stack
288GB ×16 = 4.6TB of high-bandwidth attached memory — the whole model fits, no offload.
Interposer / NVLink
NVLink-C2C fabric so the 16 GPUs act like one pool.
Per-second metering
$184/hr → 0.0511/sec. Pause the deployment and the meter hits $0.
Scale-to-zero
Autoscale + region pins (GLOBAL / US / EU / APAC) on the control plane.