NVIDIA · Hopper · 141 GB · Qwen3.8 27B · Q4_K_M at 8K context: ~180 tok/s
NVIDIA H200 141GB SXM5
The H200 as it ships in servers: 141 GB of HBM3e at 4.8 TB/s on the full 700 W Hopper chip. Usually cheaper than the PCIe card because it only fits an HGX baseboard or an adapter board. Listed because modules do surface on eBay.
No live price yet
We check eBay every 28 minutes. A price appears as soon as a listing passes our filters.
Or rent one from $3.59 an hour
Price alert
One email per move, item plus shipping. Clear the price to hear about any drop. Manage alerts.
Price history
Daily low across every tracked vendor, last 3 days.
Where to buy
| Vendor | Price | Condition | Observed | |
|---|---|---|---|---|
| eBay cheapest active listing |
$44,950 free shipping |
new and used | 2026-11-13 16:43 UTC | Go to eBay |
On eBay now
No active listing passes our filters at the moment. Checked 7 min ago. We check every 28 minutes.
Rent one by the hour
Try it before you buy, or skip buying: the cheapest on-demand price for one H200 141GB SXM5 in the cloud right now, checked every hour.
| Cloud | Per hour | Listed as | Available | |
|---|---|---|---|---|
| Runpod | $3.59 | H200 SXM | stock: low | Rent on Runpod |
| Vast.ai | $3.59 | H200 | 6 verified offers | Rent on Vast.ai |
Rental prices for every card · Rental links are referral links: we earn a small share of what you spend, at no cost to you.
What it runs
For each model, grouped by vendor with the newest first: the largest quant that fits with 8K of context and the estimated speed for one user, at 8K and at 32K of context. On a card, a model too big for its memory can still run with part of the weights in system RAM; we assume 64 GB, and the speed shown accounts for it. The model you picked on the ladder is highlighted. Estimates from the spec sheet, not measurements: see how we estimate.
| Model | Quant | Needs at 8K | Fit | Max context | Decode at 8K | At 32K | Prefill tok/s |
|---|---|---|---|---|---|---|---|
| DeepSeek | |||||||
| DeepSeek V4.1 Flash 552B-A16B Sep 2026 | IQ2_M | 189 GB | runs with 49 GB in system RAM | ~26 | ~25 | ~5,400 | |
| DeepSeek V4 Flash 284B-A13B Jul 2026 | IQ2_M | 98 GB | fits | 128K+ | ~200 | ~180 | ~13,000 |
| DeepSeek V3.1 671B-A37B Aug 2025 | Q4_K_M | 411 GB | does not fit | ||||
| Gemma | |||||||
| Gemma 4 12B May 2026 | Q8_0 | 14 GB | fits | 128K+ | ~240 | ~240 | ~14,000 |
| Gemma 4 31B Mar 2026 | Q8_0 | 35 GB | fits | 128K+ | ~94 | ~89 | ~5,600 |
| Gemma 4 26B-A4B Mar 2026 | Q8_0 | 28 GB | fits | 128K+ | ~210 | ~200 | ~46,000 |
| Gemma 3 27B Mar 2025 | Q8_0 | 31 GB | fits | 128K+ | ~100 | ~98 | ~6,300 |
| GLM | |||||||
| GLM-5.3 753B-A40B Aug 2026 | Q4_K_M | 462 GB | does not fit | ||||
| GLM-5.3 Flash 320B-A18B Aug 2026 | IQ2_M | 111 GB | fits | 128K+ | ~180 | ~170 | ~9,600 |
| GLM-4.6 355B-A32B Sep 2025 | IQ2_M | 125 GB | fits | 48K | ~130 | ~91 | ~5,400 |
| GLM-4.5 Air 106B-A12B Jul 2025 | Q8_0 | 115 GB | fits | 128K+ | ~120 | ~100 | ~14,000 |
| gpt-oss | |||||||
| gpt-oss-120b Aug 2025 | Q8_0 | 126 GB | fits | 128K+ | ~180 | ~170 | ~34,000 |
| gpt-oss-20b Aug 2025 | Q8_0 | 24 GB | fits | 128K+ | ~210 | ~190 | ~48,000 |
| Llama | |||||||
| Llama 4 Scout 109B-A17B Apr 2025 | Q8_0 | 119 GB | fits | 119K | ~100 | ~88 | ~10,000 |
| Llama 3.3 70B Dec 2024 | Q8_0 | 79 GB | fits | 128K+ | ~40 | ~36 | ~2,500 |
| Llama 3.1 8B Jul 2024 | Q8_0 | 11 GB | fits | 128K+ | ~330 | ~240 | ~22,000 |
| MiMo | |||||||
| MiMo V2.6 Pro 1.02T-A42B Sep 2026 | Q4_K_M | 624 GB | does not fit | ||||
| MiMo V2.6 Flash 309B-A15B Sep 2026 | IQ2_M | 107 GB | fits | 128K+ | ~190 | ~180 | ~12,000 |
| Mistral | |||||||
| Mistral Medium 3.5 128B Mar 2026 | Q8_0 | 140 GB | fits | 10K | ~23 | ~1,400 | |
| Mistral Small 4 119B-A6.5B Mar 2026 | Q8_0 | 132 GB | fits | 22K | ~140 | ~27,000 | |
| Mistral Small 3.2 24B Jun 2025 | Q8_0 | 28 GB | fits | 128K+ | ~120 | ~100 | ~7,200 |
| Nemotron | |||||||
| Nemotron 3.5 Lightning 30B-A3B Aug 2026 | Q8_0 | 35 GB | fits | 128K+ | ~210 | ~210 | ~49,000 |
| Qwen | |||||||
| Qwen3.8 Flash-Next 125B-A6B Aug 2026 | Q8_0 | 134 GB | fits | 128K+ | ~180 | ~170 | ~29,000 |
| Qwen3.8 27B Aug 2026 (your pick) | Q8_0 | 31 GB | fits | 128K+ | ~110 | ~100 | ~6,400 |
| Qwen3.6 27B Apr 2026 | Q8_0 | 31 GB | fits | 128K+ | ~110 | ~100 | ~6,400 |
| Qwen3.6 35B-A3B Apr 2026 | Q8_0 | 39 GB | fits | 128K+ | ~220 | ~210 | ~58,000 |
| Qwen3.5 9B Feb 2026 | Q8_0 | 11 GB | fits | 128K+ | ~320 | ~290 | ~19,000 |
| Qwen3 Coder 480B-A35B Jul 2025 | IQ2_M | 167 GB | runs with 26 GB in system RAM | ~19 | ~15 | ~2,500 | |
| Qwen3 235B-A22B Jul 2025 | Q3_K_M | 118 GB | fits | 123K | ~130 | ~110 | ~7,900 |
| Qwen3 32B Apr 2025 | Q8_0 | 38 GB | fits | 128K+ | ~85 | ~72 | ~5,300 |
Specifications
- Memory
- 141 GB HBM3e
- Memory bandwidth
- 4,800 GB/s
- Bus width
- 6144-bit
- Shader cores
- 16,896
- Tensor cores
- 528
- FP32
- 67.0 TFLOPS
- FP16 (dense)
- 989 TFLOPS
- FP8 (dense)
- 1979 TFLOPS
- Power
- 700 W
- Form factor
- SXM5 module: needs an HGX baseboard, or an SXM5-to-PCIe adapter board that supports it, and serious cooling
- Interface
- PCIe 5.0 x16
- Architecture
- Hopper
- Launched
- 2024
Source: manufacturer specifications.