NVIDIA / discrete

NVIDIA H200 SXM

141 GB decides what fits. 4800 GB/s decides how fast it runs once it does.

Quick answer

The NVIDIA H200 SXM fits 249 of 320 sized models entirely in its 141 GB; the largest widely used one is Qwen3.8-Flash-Next (111.6 GB at Q4_K_M). Memory decides what fits; its 4800 GB/s of bandwidth decides how fast it answers.

Largest popular model that fitsQwen3.8-Flash-Next180B params · Q4_K_M · needs 111.6 GB~907 tok/s Faster than you readOpen in the calculator →
Best fast pickStep-3.7-Flash201B params · Q4_K_M · needs 128.2 GB~282 tok/s Faster than you readOpen in the calculator →
Most downloaded that fitsQwen3.8-27B28B params · Q4_K_M · needs 18.6 GB~220 tok/s Faster than you readOpen in the calculator →

At 8,192 tokens of context with the whole model in device memory. Speeds are estimates from memory bandwidth, not benchmarks run on this card.

NVIDIA H200 SXM · 141 GB · 4800 GB/s · Q4_K_M where published · 8,192 tokensSpecificationpublished dataDecode speedestimate

249 of 320 fit entirely

249 of the 320 models the engine can size fit entirely in device memory at Q4_K_M where it is published, otherwise the nearest published format, and 8,192 tokens. 10 run with some layers on system memory (32 GB assumed), and 61 do not run at all.

78%of the models the engine can size fit entirely

141GB of device memory

4800GB/s memory bandwidth

10run with system memory

61do not run at 8,192 tokens

Featured models · Q4_K_M at 8,192 tokens, including offload

Model	Needs	Verdict	Decode	Calculator
Qwen3.5-2B2.3B parameters	2.32 GB	fits	~2646 tok/s	Open →
NVIDIA-Nemotron-3-Nano-4B-BF164.0B parameters	3.46 GB	fits	~1226 tok/s	Open →
Qwen3-VL-8B-Instruct8.8B parameters	7.04 GB	fits	~555 tok/s	Open →
Qwen3.5-9B9.7B parameters	7.05 GB	fits	~672 tok/s	Open →
gemma-4-12B-it12B parameters	9.54 GB	fits	~399 tok/s	Open →
gpt-oss-20b21B parameters	13.8 GB	fits	~1305 tok/s	Open →
Qwen3.8-27B28B parameters	18.6 GB	fits	~220 tok/s	Open →
Qwen3.6-27B28B parameters	18.6 GB	fits	~220 tok/s	Open →
Kimi-Linear-48B-A3B-Instruct49B parameters	30.4 GB	fits	~1746 tok/s	Open →
Qwen3.8-Flash-Next180B parameters	111.6 GB	fits	~907 tok/s	Open →
DeepSeek-V4-Flash-Vision-Exp305B parameters	188.1 GB	needs 47.9 GB of system RAM; 32 GB assumed	does not run	Open →
GLM-5.3-Flash321B parameters	198.3 GB	needs 61.4 GB of system RAM; 32 GB assumed	does not run	Open →

These examples are selected from prominent labs using the catalogue’s latest Hugging Face 30-day downloads and repository-creation freshness signal, with newer releases guaranteed a place. Offloaded rows assume 32 GB of system RAM, and a speed is only shown for a row that runs. Fitting in memory is not the same as loading: whether the runtime and version you have supports each architecture and format on this machine has not been tested here. Each “Open” link carries the same model, format, context and RAM into the calculator.

Fits entirely in NVIDIA H200 SXM memory — 249 of 320 sized models

Every model that fitswhole model resident · offload off

The most demanding model that fits is Step-3.7-Flash at Q4_K_M: 128.2 GB of the 141 GB, leaving 12.8 GB spare.

Fully resident at 8,192 tokens (or the model’s own maximum, where that is shorter), offload off, at 4800 GB/s. Each row is a run of the engine for this configuration; the rows start with current, prominent releases and “fits” is memory, not a tested runtime. Older or less prominent models remain available through this search and “Show all”.

Find a model in this list

Showing 40 of 249 models that fit.

Model	Format	Needs	Spare	Decode	Downloads / 30d	Calculator
Qwen3.8-27BQwen · 28B params	Q4_K_M	18.6 GB	122.4 GB	~220 tok/s	6.9M	Open →
Qwen3.8-Flash-NextNEWQwen · 180B params	Q4_K_M	111.6 GB	29.4 GB	~907 tok/s	1.4M	Open →
gemma-4-26B-A4B-itGoogle · 26B params	Q4_K_M	16.8 GB	124.2 GB	~1226 tok/s	13M	Open →
gemma-4-31B-itGoogle · 31B params	Q4_K_M	22.2 GB	118.8 GB	~159 tok/s	9.9M	Open →
Qwen3.5-9BQwen · 9.7B params	Q4_K_M	7.05 GB	134.0 GB	~672 tok/s	9M	Open →
Qwen3.5-4BQwen · 4.7B params	Q4_K_M	3.98 GB	137.0 GB	~1208 tok/s	7.8M	Open →
Qwen3.6-35B-A3BQwen · 36B params	Q4_K_M	23.1 GB	117.9 GB	~1760 tok/s	3.3M	Open →
gemma-4-12B-itGoogle · 12B params	Q4_K_M	9.54 GB	131.5 GB	~399 tok/s	1.9M	Open →
Qwen3.6-27BQwen · 28B params	Q4_K_M	18.6 GB	122.4 GB	~220 tok/s	2.5M	Open →
Qwen3.5-2BQwen · 2.3B params	Q4_K_M	2.32 GB	138.7 GB	~2646 tok/s	4.9M	Open →
gemma-4-E4B-itGoogle · 8.0B params	Q4_K_M	5.86 GB	135.1 GB	~1079 tok/s	4.4M	Open →
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16NVIDIA · 32B params	Q4_K_M	20.3 GB	120.7 GB	~167 tok/s	530K	Open →
NVIDIA-Nemotron-3-Nano-4B-BF16NVIDIA · 4.0B params	Q4_K_M	3.46 GB	137.5 GB	~1226 tok/s	3.4M	Open →
gemma-4-E2B-itGoogle · 5.1B params	Q4_K_M	4.02 GB	137.0 GB	~2850 tok/s	3M	Open →
Qwen3.5-0.8BQwen · 873M params	Q4_K_M	1.46 GB	139.5 GB	~5796 tok/s	2.6M	Open →
Qwen3-VL-8B-InstructQwen · 8.8B params	Q4_K_M	7.04 GB	134.0 GB	~555 tok/s	14.6M	Open →
Qwen3.5-27BQwen · 28B params	Q4_K_M	18.6 GB	122.4 GB	~220 tok/s	1.9M	Open →
North-Micro-Vision-InstructCohereLabs · 2.5B params	Q4_K_M	2.99 GB	138.0 GB	~1740 tok/s	180.7K	Open →
Qwen3.5-35B-A3BQwen · 36B params	Q4_K_M	23.1 GB	117.9 GB	~1760 tok/s	1.6M	Open →
NVIDIA-Nemotron-3-Super-120B-A12B-BF16NVIDIA · 124B params	Q4_K_M	76.9 GB	64.1 GB	~43 tok/s	1.2M	Open →
granite-4.2-8bIBM · 8.8B params	Q4_K_M	7.49 GB	133.5 GB	~505 tok/s	128.4K	Open →
GLM-4.7-Flashzai-org · 31B params	Q4_K_M	20.4 GB	120.6 GB	~166 tok/s	1.8M	Open →
granite-4.1-3bIBM · 3.4B params	Q4_K_M	3.72 GB	137.3 GB	~1178 tok/s	521.8K	Open →
LFM2.5-2.6BLiquidAI · 2.7B params	Q4_K_M	2.61 GB	138.4 GB	~1890 tok/s	108.1K	Open →
Qwen3-VL-4B-InstructQwen · 4.4B params	Q4_K_M	4.72 GB	136.3 GB	~881 tok/s	3.5M	Open →
granite-4.1-30bIBM · 29B params	Q4_K_M	20.4 GB	120.6 GB	~166 tok/s	301.5K	Open →
Qwen3.5-122B-A10BQwen · 125B params	Q4_K_M	77.9 GB	63.1 GB	~644 tok/s	512.6K	Open →
Qwen3-VL-2B-InstructQwen · 2.1B params	Q4_K_M	3.05 GB	138.0 GB	~1598 tok/s	2.8M	Open →
granite-4.2-3bIBM · 3.7B params	Q4_K_M	3.72 GB	137.3 GB	~1178 tok/s	50.6K	Open →
LFM2.5-230MLiquidAI · 230M params	Q4_K_M	1.05 GB	139.9 GB	~13174 tok/s	88.6K	Open →
Qwen3-0.6BQwen · 752M params	Q4_K_M	2.22 GB	138.8 GB	~2451 tok/s	29.7M	Open →
Qwen3-Coder-NextQwen · 80B params	Q4_K_M	50.0 GB	91.0 GB	~1171 tok/s	596.3K	Open →
gpt-oss-20bOpenAI · 21B params	Q4_K_M	13.8 GB	127.2 GB	~1305 tok/s	6.6M	Open →
granite-4.2-30bIBM · 29B params	Q4_K_M	20.7 GB	120.3 GB	~166 tok/s	35.8K	Open →
granite-4.1-8bIBM · 8.8B params	Q4_K_M	7.49 GB	133.5 GB	~505 tok/s	179.3K	Open →
NVIDIA-Nemotron-3-Nano-30B-A3B-BF16NVIDIA · 32B params	Q4_K_M	20.3 GB	120.7 GB	~167 tok/s	875.2K	Open →
gpt-oss-120bOpenAI · 117B params	Q4_K_M	71.9 GB	69.1 GB	~918 tok/s	4.5M	Open →
LFM2.5-VL-3BLiquidAI · 3.1B params	Q4_K_M	2.85 GB	138.1 GB	~1890 tok/s	26.7K	Open →
MiMo-V2.6-Distill-Qwen-9BNEWXiaomiMiMo · 9.4B params	Q4_K_M	6.90 GB	134.1 GB	~672 tok/s	14.6K	Open →
Qwen3-4B-Instruct-2507Qwen · 4.0B params	Q4_K_M	4.72 GB	136.3 GB	~881 tok/s	3.7M	Open →

Sized at the format most people actually download, not at FP16. The catalogue ordered by downloads →

Rent it instead

Rent an H200 by the hour

What the NVIDIA H200 SXM costs on Vast.ai and RunPod right now, per machine, from one card to eight. Every price is read from the provider's own API.

Prices · 18:30 UTC, 22 Sept

1× H200 · 141 GB

$3.59per hour

RunPod · Community CloudRunPod $4.59/h Secure CloudRent on RunPod

2× H200 · 282 GB

$9.18per hour

RunPod · Secure CloudVast.ai $10.00/h Rent on RunPod

4× H200 · 564 GB

$18.36per hour

RunPod · Secure CloudVast.ai $18.42/h Rent on RunPod

8× H200 · 1128 GB

$26.52per hour

Vast.ai · Marketplace · 99.3% reliableRent on Vast.ai

Buy one or rent one?

Your price, your hours, your electricity. Everything else is arithmetic.

Price you would pay ($)

Resale after 3 years ($)

Hours of use per day4 h

$ per kWh

Watts

Rent $/h

Watts start at the NVIDIA H200 SXM’s published board power (700 W), an upper bound: decoding rarely holds a card at its limit. Rent starts at the cheapest live price (RunPod). The electricity price is a placeholder — put yours in.

At 4 h a day, renting costs $5,241.40 a year.

That is $436.78 a month, and nothing when the machine is stopped.
Owning would cost $255.50 a year in electricity, on top of the price.
Every $1,000 of purchase price takes 0.2 years of this use to earn back.

Enter the price you would pay to see the exact break-even.

On-demand prices for the whole machine. Vast hosts below 98% measured reliability are left out; RunPod’s Community Cloud is vetted third-party hosts and its Secure Cloud is data-centre capacity. Every rentable card compared.

Referral links Vast.ai, RunPod and Novita pay us a share of what you spend if you sign up through these buttons. It costs you nothing, and it never decides an order or a recommendation: both are computed from the live price and the speed, and options that pay us nothing are listed and recommended on the same terms. How we rank

Under the hood

The specification behind every figure

What the manufacturer publishes for this device, and the pages it was read from.

Manufacturer specification

Memory scope	dedicated
Capacity	141 GB
Published options	141 GB
Bandwidth	4800 GB/s
Memory type	HBM3e
Bus width	Not published
FP32 peak	67 TFLOPS
Dense matrix peak	Not published without sparsity
Max thermal design power (TDP), configurable ceiling	700 W

Source ledger

NVIDIA H200 GPU ↗

Retrieved 2026-08-30

Caveats

NVIDIA marks the H200 table specifications as preliminary and subject to change.
No dense matrix throughput is catalogued for this device: NVIDIA's H200 page publishes "FP16 Tensor Core 1,979 teraFLOPS" footnoted "With sparsity", and no dense figure. Time to first token is withheld rather than priced against a sparsity figure.

Published capacity is a hardware ceiling, not guaranteed free runtime memory. The calculator shows the runtime reserve separately rather than folding it into a single number.

Run the diagnostic on the NVIDIA H200 SXM →