MI300X
The MI300X is a cloud GPU with 192 GB of memory and 750 W TDP. It is currently available from 3 providers starting at $0.50/hr, with a market median of $2.39/hr across 7 offerings.
Best deal
Hardware specifications
- Architecture
- CDNA 3
- VRAM
- 192 GB
- Memory type
- HBM3
- Memory bandwidth
- 5,200 GB/s
- FP16
- 1307 TFLOPS
- FP32
- 163 TFLOPS
- TDP
- 750 W
Same across all providers.
Market comparison
Offerings
LLMs that fit
| Model | Params | 16-bit | 8-bit | 4-bit |
|---|---|---|---|---|
DeepSeek V4 Pro DeepSeek | 1.6T 49B active | Not released | Not released | Barely fits on 7× |
Kimi K2.5 Moonshot AI | 1T 32B active | Not released | Not released | Barely fits on 5× |
GLM-5.2 Z.ai | 753B | Needs >8 GPUs | Barely fits on 7× | Barely fits on 4× |
DeepSeek R1 DeepSeek | 671B 37B active | Not released | 6× · 14% spare | 4× · 190.8 GB free |
DeepSeek V4 Flash DeepSeek | 284B | Not released | Not released | 2× · 153.6 GB free |
Solar Open2 250B Upstage | 250B 15B active | Barely fits on 4× | 3× · 180 GB free | 2× · 153.6 GB free |
Qwen3 235B-A22B Alibaba | 235B 22B active | Barely fits on 4× | Barely fits on 2× | 2× · 166.8 GB free |
Laguna-S 2.1 Poolside | 118B 8B active | Barely fits on 2× | Barely fits on 1× | 1× · 84 GB free |
gpt-oss-120b OpenAI | 117B 5.1B active | Not released | Not released | 1× · 98.4 GB free |
Llama 3.1 70B Meta | 70B | Barely fits on 1× | 1× · 100.2 GB free | 1× · 138 GB free |
Qwen3.6 35B-A3B Alibaba | 35B 3B active | 1× · 88.8 GB free | 1× · 135.6 GB free | 1× · 158.4 GB free |
Qwen3 32B Alibaba | 32.8B | 1× · 97.2 GB free | 1× · 140.4 GB free | 1× · 162 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
No similar GPUs found in our catalog.
See more GPUs like this in
Frequently asked questions
How much does the MI300X cost per hour?
The cheapest MI300X offering starts at $0.50/hr, and the market median is $2.39/hr.
Which cloud providers offer the MI300X?
We track MI300X offerings from HardwareHQ, RunPod, GPU Finder.
How much VRAM does the MI300X have?
The MI300X has 192 GB of VRAM.
What LLMs can run on the MI300X?
With 192 GB of VRAM, the MI300X can run 16 of our tracked models in 4-bit quantization, including DeepSeek V4 Pro, Kimi K2.5, GLM-5.2. 10 models fit in 16-bit precision.
Where is the MI300X cheapest?
The cheapest MI300X offering in our catalog is HardwareHQ at $0.50/hr per GPU-hour (×1).