datacenter
Rent AMD MI300X.
CDNA 3
192GB VRAM
From $0.50/hr
Per hour
$0.50
Per day
$12.00
Per week
$84.00
Per month
$360
Price history
Daily median across providers.
Loading...
Where to rent it
All providers carrying this GPU.
FAQ
Frequently asked.
How much does it cost to rent the AMD MI300X right now?
The cheapest current listing is $0.50/hr on RunPod, the only provider listing it right now. Running one around the clock for a month (730 hours) at the cheapest rate costs about $365.
What AI models can the AMD MI300X run?
With 192 GB it holds a 70B model at INT8/FP8 (about 70 GB of weights) on one card, or at 4-bit with plenty of room for long contexts and many concurrent requests. It can QLoRA fine-tune 70B models. FP16 70B (about 140 GB) needs two cards.
Which models are recommended for the AMD MI300X?
6 models in our catalog have a recommended configuration for this card, with the quantization that fits its 192 GB — see the AI models tab.
Why does the AMD MI300X cost different amounts on different providers?
Marketplaces such as Vast.ai and Clore.ai list machines from independent hosts who set their own prices, so competition pushes rates down but reliability varies by host. Operator clouds such as Lambda and the hyperscalers charge more for consistent hardware, support and networking. Prices here are per GPU, so they compare directly.
How is the AMD MI300X price calculated?
We collect listings from each provider's public API or price page — every 5–10 minutes for marketplaces, up to daily for operator clouds — and show the median hourly price per provider from the last 24 hours, so one unusually cheap listing doesn't set the headline. Details are on the methodology page.
AI models
INT4
INT4
INT4
INT4
INT4
INT4
Models that run on this GPU.
GPU-count + quantization recommendations covering fine-tuning, inference, and run-it-yourself scenarios on the AMD MI300X.
Baidu: ERNIE 4.5 300B A47B
by Baidu
Required GPUs
1× AMD MI300X
Total VRAM
192 GB
Parameters
300B
Context
131K tokens
DeepSeek Coder V2 236B
by DeepSeek
Required GPUs
1× AMD MI300X
Total VRAM
192 GB
Parameters
236B (21B active)
Context
128K tokens
Qwen3 235B A22B Instruct 2507 FP8 Throughput
by Alibaba (Qwen Team)
Required GPUs
1× AMD MI300X
Total VRAM
192 GB
Parameters
235B
Context
262K tokens
Qwen3-VL-235B-A22B-Instruct-FP8
by Alibaba (Qwen Team)
Required GPUs
1× AMD MI300X
Total VRAM
192 GB
Parameters
235B
Context
262K tokens
Qwen 3 235B
by Alibaba (Qwen Team)
Required GPUs
1× AMD MI300X
Total VRAM
192 GB
Parameters
235B (22B active)
Context
128K tokens
Qwen: Qwen3 Max
by Alibaba (Qwen Team)
Required GPUs
1× AMD MI300X
Total VRAM
192 GB
Parameters
235B
Context
262K tokens
Renting for inference?
Pair these models with the cheapest provider on the rental table.
See rental rates on the Overview tab →
FAQ
AI models on this GPU.
How much does it cost to rent the AMD MI300X right now?
The cheapest current listing is $0.50/hr on RunPod, the only provider listing it right now. Running one around the clock for a month (730 hours) at the cheapest rate costs about $365.
What AI models can the AMD MI300X run?
With 192 GB it holds a 70B model at INT8/FP8 (about 70 GB of weights) on one card, or at 4-bit with plenty of room for long contexts and many concurrent requests. It can QLoRA fine-tune 70B models. FP16 70B (about 140 GB) needs two cards.
Which models are recommended for the AMD MI300X?
6 models in our catalog have a recommended configuration for this card, with the quantization that fits its 192 GB — see the AI models tab.
Why does the AMD MI300X cost different amounts on different providers?
Marketplaces such as Vast.ai and Clore.ai list machines from independent hosts who set their own prices, so competition pushes rates down but reliability varies by host. Operator clouds such as Lambda and the hyperscalers charge more for consistent hardware, support and networking. Prices here are per GPU, so they compare directly.
How is the AMD MI300X price calculated?
We collect listings from each provider's public API or price page — every 5–10 minutes for marketplaces, up to daily for operator clouds — and show the median hourly price per provider from the last 24 hours, so one unusually cheap listing doesn't set the headline. Details are on the methodology page.