H20
The H20 is a cloud GPU with 141 GB of memory. It is currently available from 1 provider starting at $1.04/hr, with a market median of $1.04/hr across 1 offering.
Hardware specifications
- Architecture
- Hopper
- VRAM
- 141 GB
- Memory type
- HBM3
Same across all providers.
Market comparison
Offerings
No offerings found
Try clearing the spot filter or check back later.
LLMs that fit
| Model | Params | 16-bit | 8-bit | 4-bit |
|---|---|---|---|---|
Kimi K2.5 Moonshot AI | 1T 32B active | Not released | Not released | 7× · 13.2% spare |
GLM-5.2 Z.ai | 753B | Needs >8 GPUs | Needs >8 GPUs | Barely fits on 5× |
DeepSeek R1 DeepSeek | 671B 37B active | Not released | Barely fits on 8× | 5× · 127.8 GB free |
DeepSeek V4 Flash DeepSeek | 284B | Not released | Not released | 2× · 51.6 GB free |
Solar Open2 250B Upstage | 250B 15B active | 6× · 14.8% spare | Barely fits on 3× | 2× · 51.6 GB free |
Qwen3 235B-A22B Alibaba | 235B 22B active | Barely fits on 5× | Barely fits on 3× | 2× · 64.8 GB free |
Laguna-S 2.1 Poolside | 118B 8B active | 3× · 84.6 GB free | 2× · 96 GB free | 1× · 33 GB free |
gpt-oss-120b OpenAI | 117B 5.1B active | Not released | Not released | 1× · 47.4 GB free |
Llama 3.1 70B Meta | 70B | 2× · 114 GB free | 1× · 49.2 GB free | 1× · 87 GB free |
Qwen3.6 35B-A3B Alibaba | 35B 3B active | 1× · 37.8 GB free | 1× · 84.6 GB free | 1× · 107.4 GB free |
Qwen3 32B Alibaba | 32.8B | 1× · 46.2 GB free | 1× · 89.4 GB free | 1× · 111 GB free |
Qwen2.5 14B Alibaba | 14B | 1× · 105.6 GB free | 1× · 121.7 GB free | 1× · 129.6 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.
Similar GPUs
See more GPUs like this in
Frequently asked questions
How much does the H20 cost per hour?
The cheapest H20 offering starts at $1.04/hr, and the market median is $1.04/hr.
Which cloud providers offer the H20?
No providers currently list the H20 in our catalog.
How much VRAM does the H20 have?
The H20 has 141 GB of VRAM.
What LLMs can run on the H20?
With 141 GB of VRAM, the H20 can run 15 of our tracked models in 4-bit quantization, including Kimi K2.5, GLM-5.2, DeepSeek R1. 10 models fit in 16-bit precision.
What's the biggest LLM I can run on multiple H20s?
With 8× H20 (1128 GB total), the largest tracked model that fits in 16-bit is Solar Open2 250B (250B params), which needs 6× GPUs.