NVIDIA GH200

96 GB VRAM989 FP16 TFLOPS700 W TDP1 providers1 offerings1 regions

The NVIDIA GH200 is an enterprise-class GPU built on the Hopper architecture, featuring 96 GB of memory with 4,000 GB/s of memory bandwidth and a 700 W TDP. It delivers 67 TFLOPS of FP32 performance and 989 TFLOPS of FP16 tensor performance. Currently available from 1 provider starting at $2.52/GPU/hour, with a market median of $2.52/GPU/hour across 1 configurations.

Cheapest

$2.520/hr

Runcrate

Median

$2.520/hr

across 1 offerings

Most expensive

$2.520/hr

Runcrate

90-day trend

+0.2%

median rate

Hardware specifications

Same across all providers

Memory

96 GB

HBM3

What LLMs fit? ↓

FP16 performance

989 TFLOPS

peak

TDP

700 W

Memory bandwidth

4,000 GB/s

Architecture

Hopper

FP32 performance

67 TFLOPS

Brand

NVIDIA

Series

GH200

Market price history

Median price across all providers

Prices updated 7 hours ago

3 offerings from 1 providers

Sorted by price ascending. Compare configurations side-by-side.

Provider	Count	vCPU	RAM	Region	Per GPU hour	Total/hr	Action
Lambda	×1	64	432 GB	--	$2.290 0.0% 30d	$2.29
RuncrateCheapest	×1	64	432 GB		$2.520 0.0% 30d	$2.52	Launch
Sesterce CloudReferral link	×1	144	480 GB	--	$4.653	$4.65

LLMs that fit on the NVIDIA GH200

Estimated VRAM for popular open-weight LLMs at 16-bit (FP16/BF16), 8-bit (FP8/INT8) and 4-bit (INT4/MXFP4) precision, against 96 GB per GPU. Showing the 12 largest that fit.

Model	Parameters	16-bit	8-bit	4-bit
Laguna-S 2.1 Poolside	118B 8B active	282 GB Too large	155 GB Too large	90 GB 1× · 5.7 GB free
gpt-oss-120b OpenAI	117B 5.1B active	— Not released	— Not released	78 GB 1× · 18 GB free
Qwen3.6 35B-A3B Alibaba	35B 3B active	86 GB 1× · 9.7 GB free	47 GB 1× · 49 GB free	28 GB 1× · 68 GB free
Qwen3 32B Alibaba	32.8B	79 GB 1× · 17 GB free	43 GB 1× · 53 GB free	25 GB 1× · 71 GB free
Gemma 4 31B Google	31.3B	75 GB 1× · 21 GB free	41 GB 1× · 55 GB free	24 GB 1× · 72 GB free
GLM-4.7 Flash Z.ai	31.2B	75 GB 1× · 21 GB free	41 GB 1× · 55 GB free	24 GB 1× · 72 GB free
Qwen3.6 27B Alibaba	27.8B	67 GB 1× · 29 GB free	37 GB 1× · 59 GB free	21 GB 1× · 75 GB free
Mistral Small 3.2 24B Mistral AI	24B	58 GB 1× · 38 GB free	32 GB 1× · 64 GB free	18 GB 1× · 78 GB free
gpt-oss-20b OpenAI	21B 3.6B active	— Not released	— Not released	17 GB 1× · 79 GB free
Gemma 3 12B Google	12.2B	29 GB 1× · 67 GB free	16 GB 1× · 80 GB free	9.4 GB 1× · 87 GB free
Granite 4.1 8B IBM	8.8B	21 GB 1× · 75 GB free	12 GB 1× · 84 GB free	6.8 GB 1× · 89 GB free
Qwen3 8B Alibaba	8.2B	20 GB 1× · 76 GB free	11 GB 1× · 85 GB free	6.3 GB 1× · 90 GB free

Estimates: model weights plus 20% for KV cache and runtime overhead, at short context lengths. Long contexts and large batches need more. Rows marked “tight” hold the weights but not that full margin. A ×N figure is the total VRAM across N of these GPUs and makes no claim about interconnect throughput. How we estimate this · Start from a model instead

See more GPUs like this in

Highest performance

Want more from this page?

See incorrect data?

Frequently asked questions

How much does the GH200 cost per hour?

The GH200 starts at $2.52 per GPU-hour, with a median of $2.52 across 1 available configurations. Actual pricing depends on the provider, region, and whether you rent on-demand, spot or reserved.

Which cloud providers offer the GH200?

3 providers currently list the GH200: Sesterce Cloud, Runcrate, Lambda. Compare their hourly pricing, regions and machine specs in the table above.

How much VRAM does the GH200 have?

The GH200 comes with 96 GB of VRAM, which determines the largest models and batch sizes it can hold in memory for training and inference.

Which LLMs can I run on the GH200?

With 96 GB of VRAM, a single GH200 can serve models such as Laguna-S 2.1, gpt-oss-120b, Qwen3.6 35B-A3B, Qwen3 32B. The table above lists the estimated VRAM for each model at 16-bit, 8-bit and 4-bit precision, including the multi-GPU configurations needed for larger models.

Where is the GH200 cheapest?

The lowest on-demand price we currently track for the GH200 is $2.52 per GPU-hour at Runcrate. Because providers update pricing and availability frequently, check the live table above before you launch.