Tool · Run a model

Will my model fit on this GPU?

Compute the VRAM you need for a given AI model at a chosen precision and context window, and see exactly which rentable GPUs have the headroom — plus the cheapest provider for each.

granite-4.2-30b
1,024 1,000,000
1 64
Params in memory
30.0B
Dense
Bytes / param
2.0
FP16
Context
131,072
tokens
VRAM required
77 GB
weights + KV + 20% headroom

GPUs that fit

Sorted by VRAM ascending — smallest fitting card first (usually the cheapest).

GPU VRAM FP16 TFLOPS TDP Cheapest provider $/hr
80GB — — Nosana $1.19/hr Open →
80GB — — Thunder Compute $3.20/hr Open →
80GB — 300W Vast.ai $0.74/hr Open →
80GB — — io.net $1.00/hr Open →
80GB — — Thunder Compute $1.09/hr Open →
80GB — — Lium $2.00/hr Open →
96GB — 1000W no live offers — Open →
96GB — 600W Cyfuture AI $1.06/hr Open →
96GB — — no live offers — Open →
96GB — — Nosana $0.91/hr Open →
96GB — 600W Hyperstack $1.85/hr Open →
96GB — — RunPod $1.69/hr Open →
96GB — — RunPod $1.69/hr Open →
96GB — — Clore.ai $0.88/hr Open →
96GB — — Vast.ai $1.47/hr Open →
96GB — — no live offers — Open →
96GB — — Vast.ai $1.31/hr Open →
141GB — — RunPod $3.59/hr Open →
141GB — — Vast.ai $3.95/hr Open →
141GB — — RunPod $0.50/hr Open →
192GB — 750W no live offers — Open →
192GB — — DeepInfra $3.69/hr Open →
192GB — — RunPod $0.50/hr Open →
256GB — 1000W no live offers — Open →
288GB — 2700W no live offers — Open →
288GB — — Lium $8.00/hr Open →
288GB — — io.net $6.73/hr Open →
288GB — 1400W no live offers — Open →

Related tools