GPU guide

H200 141GB for AI: what it actually runs

By Bitpute · Published 12 July 2026 · Updated 24 July 2026 · Specs verified against manufacturer datasheets 24 July 2026 · How we estimate · Sources · Editorial policy · Version history · Report an error · NVIDIA H200 datasheet

Form factor: SXM. The figures on this page are for the H200 SXM module: 4.8 TB/s and 700 W. The NVL variant shares the same 141 GB and 4.8 TB/s but drops to a 600 W TDP. Source: NVIDIA datasheet.

AI suitability

That is its whole pitch: jobs that were multi-GPU problems become single-card problems, which simplifies deployment more than any benchmark shows. For everything 32B and below it is overkill you pay for by the hour.

For LLM inference the number that matters most is memory bandwidth, because generating each token means reading the whole model. On the H200 141GB that puts the theoretical ceiling around 986 tokens/second on Llama 3.1 8B at Q4_K_M, and about 109 tokens/second on the biggest model it comfortably holds, Qwen2.5 72B — real-world throughput lands below these ceilings. At Q4_K_M with an 8K context this card comfortably fits 12 of the 12 models in our database; at the near-lossless Q8_0, 12.

Which models fit the H200 141GB?

Computed at an 8K context (or the model's own cap). ✓ fits comfortably (≤95% of 141 GB) · ~ tight · ✗ doesn't fit. Every model links to its own guide.

Electricity

The board is rated at 700 W. Run it under sustained load 8 hours a day and that's about 168 kWh a month — roughly $20/month at an example rate of $0.12/kWh (set your own tariff in the workspace). Idle and light chat draw far less; the figure above is the worst case, not the typical bill.

Renting instead of buying

This is a cloud-native card — the H200 141GB is something you rent far more often than you buy. Hourly rates move constantly, so we don't print them here; the workspace holds the current figures and shows rental cost against electricity for any model you pick.

Alternatives

Compare any two of these head-to-head — speed on the same model, cost, power — in GPU Compare.

Why bandwidth matters more than core count for token generation: memory bandwidth explained.

Capacity against speed

The two axes are independent. Position on the horizontal decides which models fit; position on the vertical decides how fast they decode once they do.

If a model will not load on this card

In rough order of likelihood. For reference, the largest model in our catalogue that fits this card at Q4_K_M is Falcon 180B at 107.5 GB.

1. The model is larger than usable VRAM, not nameplate VRAM. This card reports 141 GB but roughly 133.9 GB is available to a model after the display buffer, driver and CUDA context. A model sized against the nameplate figure will appear to fit and then fail.
2. KV cache grew past the headroom. Weights are fixed; the cache is not. A load that succeeds at 2K context can fail at 32K on the same card, because the cache comes out of the same budget. If it loaded yesterday and fails today, context length is the first thing to check. See KV cache.
3. The quantization is heavier than assumed. Q4_K_M is roughly 4.85 bits per weight, not 4. On a large model that difference is gigabytes. Confirm which file you actually downloaded — see which quantization to run.
4. Something else is already holding VRAM. A browser with hardware acceleration, another model still resident, or a previous process that did not release memory. On Linux nvidia-smi shows what is allocated.
5. It loaded but generation is very slow. That usually means layers were offloaded to system RAM rather than the load failing outright. The model runs, but every offloaded layer crosses PCIe on each token. Reduce context, drop a quantization level, or use a card with more VRAM.

Common questions

Is the H200 141GB good for AI and local LLMs?

Yes, within its tier: with 141 GB of VRAM and 4800 GB/s of memory bandwidth it comfortably runs 12 of the 12 models in our database at Q4_K_M with an 8K context, the largest being Qwen2.5 72B. Memory bandwidth caps generation speed at roughly 986 tokens/second on an 8B model in theory, with real-world results lower.

What is the biggest model an H200 141GB can run?

Qwen2.5 72B at Q4_K_M with an 8K context is the largest comfortable fit (46.3 GB of 141 GB). Larger models need a bigger card or a multi-GPU split.

How much electricity does an H200 141GB use?

The board is rated at 700 W. Under sustained load for 8 hours a day that is about 168 kWh a month — roughly $20/month at an example rate of $0.12/kWh. Idle draw is far lower, so light interactive use costs much less.

Data centre cards, ranked by VRAM

Server GPUs: HBM bandwidth, NVLink, MIG partitioning, rack thermals. Every card in this tier, smallest memory first — the one you are reading is highlighted. Cards shown without a memory figure are not in our calculation engine yet.

Where it sits in the stack

The H200 141GB is a Hopper-generation part with HBM3e memory and 4th-generation Tensor Cores. Memory type is why the bandwidth figure above lands where it does: HBM stacks feed far more bandwidth per watt than GDDR, and decode speed on a model that already fits is set by bandwidth, not by core count.

For multi-GPU work it supports NVLink — 900 GB/s between GPUs — which matters when a model is split across cards, because tensor-parallel inference moves activations between GPUs on every token.

Its Tensor Cores handle FP16, BF16, INT8 and FP8. FP8 halves activation memory against FP16 where a runtime supports it; Ampere cards do not have it. Most local inference runs weights at 4-6 bits via GGUF quantization rather than a native Tensor Core format, so these matter most for training and for server runtimes such as TensorRT-LLM and vLLM.

Hopper adds the Transformer Engine and FP8, which cuts activation memory and speeds up attention on supported runtimes. MIG can slice it into as many as seven isolated instances, so one board can serve several small models at once. None of this changes the VRAM arithmetic on this page — weights and KV cache still have to fit.

On the software side it is a CUDA device like any other NVIDIA card, so PyTorch, TensorRT, vLLM, llama.cpp and Ollama all run on it unchanged. What differs between cards is not compatibility but how much fits and how fast it decodes — which is what the numbers above measure. For the quantization formats these runtimes expect, and how much room the GPU Memory Calculator says you have left, start there.

Head to head

Pre-computed comparisons against the cards it is usually weighed against:

H100 80GB vs H200 141GB