LLM Opex

Models

NVIDIA-Nemotron-3-Ultra-550B-A55B-FP8-dynamic: GPU memory and hosting cost

Open weights · 560.52 billion parameters · licence other

No cloud sells this model per token: to use it you run it yourself, on GPU instances you rent.

Run it yourself

It needs 700.6 GB of GPU memory at fp8, the precision it ships in.

CloudCheapest instance that fitsGPUsAn hourA month, around the clock
Crusoemi300x.8x8 × MI300X 192 GB$27.60$20,148
Nebiusgpu-h200-sxm.8gpu-128vcpu-1600gb8 × H200 141 GB$43.20$31,536
Together AIhgx-h200.8gpu8 × H200 141 GB$47.92$34,982
OCIBM.GPU.MI300X.88 × MI300X 192 GB$48.00$35,040
AWS2 × p4de.24xlarge, joined by EFA8 × A100 80 GB in each$54.89$40,073
Alibaba Cloudecs.gn8v-8x.16xlarge8 × GPU H 96 GB$58.49$42,698
Azure2 × Standard_ND96amsr_A100_v4, joined by InfiniBand8 × A100 80 GB in each$65.54$47,844
Google Clouda3-ultragpu-8g8 × H200 141 GB$84.81$61,909

Several servers: the model is sized across identical servers that hold it together, on instance types whose vendor documents a network between servers; the price is all of them. It is shown where no one server holds the model, or where it costs less than the one that does.

More from Other

Similar size, open weights

List prices collected 21 November 2026, refreshed every three hours. No negotiated rates or taxes.