grab▲gpu

GPU model · price + stock · hourly snapshot

Nvidia GB200 cloud rental prices

1 provider lists the Nvidia GB200, 1 has it in stock. It rents for $8.13 per GPU hour.

Grace Blackwell 186 GB VRAM Q1 2024 2 configurations

LIVE Prices and stock checked 1 minute ago

Cheapest in stock
$8.13
Vast.ai
Median
$8.13
across 1 provider
Most expensive
$8.13
the only price listed

Providers

Every provider renting the Nvidia GB200

All providers →

Billing:

On-demandfrom $8.13/hr1 provider in stockPay as you go, stop any time.Spotfrom $7.50/hr1 provider in stockCheaper, but the provider can take the machine back.Reservednone listedPrepay a term for the lowest hourly rate.
Providers renting the Nvidia GB200, in stock first, then cheapest first
Provider Country Configs Scale From Stock Days in stock Grab
Vast.ai United States of America 2 2x-4x $8.13 per sec In stock 2/2 Grab

In stock first, then cheapest. Stock is what the provider reported at our last check. One GPU for a month (730h) at $8.13: $5,935.

Price history

Price trend

Cheapest Nvidia GB200 rate has held at $8.13/GPU/hr since Nov 3.

$8.13 /GPU/hr · Nov 4

One point per day: the lowest listed $/GPU/hr across every provider we track, captured daily. 19 offers in the latest snapshot.

Configurations

All 2 configurations

2 of 2 configurations
Nvidia GB200 configurations from every provider
ConfigurationProviderRegionVRAMvCPURAMPer GPU/hrTotal/hrStock
4x GB200SpotVast.aiVast.aiID740 GB144957 GB$7.50per sec$30.01In stock
2x GB200Vast.aiVast.aiID370 GB72478 GB$8.13per sec$16.26In stock

Model fit

What fits in 186 GB

The question behind most rentals: will the model load. Sized against one Nvidia GB200.

VRAM required to serve open-weight LLMs on a Nvidia GB200, by quantization precision
Model 16-bit8-bit4-bit
DeepSeek-V3 DeepSeek · 671B params · 37B active 1,610 GB 9× · 64 GB free 805 GB 5× · 125 GB free 403 GB 3× · 155 GB free
Llama 3.1 405B Meta · 405B params 972 GB 6× · 144 GB free 486 GB 3× · 72 GB free 243 GB 2× · 129 GB free
Qwen3 235B-A22B Alibaba · 235B params · 22B active 564 GB 4× · 180 GB free 282 GB 2× · 90 GB free 141 GB 1× · 45 GB free
Mixtral 8x22B Mistral · 141B params · 39B active 338 GB 2× · 34 GB free 169 GB 1× · 17 GB free 85 GB 1× · 101 GB free
Mistral Large 2 Mistral · 123B params 295 GB 2× · 77 GB free 148 GB 1× · 38 GB free 74 GB 1× · 112 GB free
gpt-oss-120b OpenAI · 117B params · 5.1B active 281 GB 2× · 91 GB free 140 GB 1× · 46 GB free 70 GB 1× · 116 GB free
Qwen2.5 72B Alibaba · 72.7B params 174 GB 1× · 12 GB free 87 GB 1× · 99 GB free 44 GB 1× · 142 GB free
Llama 3.3 70B Meta · 70.6B params 169 GB 1× · 17 GB free 85 GB 1× · 101 GB free 42 GB 1× · 144 GB free
Qwen3 32B Alibaba · 32.8B params 79 GB 1× · 107 GB free 39 GB 1× · 147 GB free 20 GB 1× · 166 GB free
Gemma 2 27B Google · 27.2B params 65 GB 1× · 121 GB free 33 GB 1× · 153 GB free 16 GB 1× · 170 GB free
Mistral Small 3 Mistral · 24B params 58 GB 1× · 128 GB free 29 GB 1× · 157 GB free 14 GB 1× · 172 GB free
Llama 3.1 8B Meta · 8B params 19 GB 1× · 167 GB free 10 GB 1× · 176 GB free 5 GB 1× · 181 GB free

Required memory is weights plus 20% for KV cache and runtime at short context; long contexts and large batches need more. Cells tinted green fit on a single Nvidia GB200.

Specifications

Nvidia GB200 specifications

Nvidia GB200 hardware specifications
ArchitectureGrace Blackwell
Memory per GPU186 GB HBM3e
Memory bandwidth8,000 GB/s
Release dateQ1 2024
GPU interconnectNVLink 1,800 GB/s
FP4 compute rate10,000 TFLOPS (dense), 20,000 with sparsity
FP8 compute rate5,000 TFLOPS (dense), 10,000 with sparsity
FP16 / BF16 compute rate2,500 TFLOPS (dense), 5,000 with sparsity
INT8 compute rate5,000 TOPS (dense), 10,000 with sparsity
Process nodeTSMC 4NP
Board power1,200 W
PartitioningMIG, up to 7 instances

Alternatives

Alternatives to the Nvidia GB200