Instinct MI250X
Best local AI models for this GPU
VRAM
128 GB
Bandwidth
3277 GB/s
Vendor
AMD
Compatible Models
29
Where to Buy
Top Compatible Models
Top 10 models ranked by performance. Click any model to expand details or view full compatibility.
Memory Usage94.1 GB
Weights
KV Cache
Overhead
Context Length32,768 tokens
Install with Ollama:
ollama run gpt-oss:120b-q5_k_m
Memory Usage44.3 GB
Weights
KV Cache
Overhead
Context Length163,840 tokens
Install with Ollama:
ollama run deepseek-r1:70b
Memory Usage44.3 GB
Weights
KV Cache
Overhead
Context Length131,072 tokens
Install with Ollama:
ollama run llama3.3:70b
Memory Usage45.5 GB
Weights
KV Cache
Overhead
Context Length131,072 tokens
Install with Ollama:
ollama run qwen2.5:72b
Memory Usage44.3 GB
Weights
KV Cache
Overhead
Context Length131,072 tokens
Install with Ollama:
ollama run llama3.1:70b
Memory Usage29.6 GB
Weights
KV Cache
Overhead
Context Length131,072 tokens
Install with Ollama:
ollama run qwen3:32b-q6_k
Memory Usage29.1 GB
Weights
KV Cache
Overhead
Context Length16,384 tokens
Install with Ollama:
ollama run gemma3:27b-q6_k
Memory Usage29.6 GB
Weights
KV Cache
Overhead
Context Length131,072 tokens
Install with Ollama:
ollama run qwen2.5:32b-q6_k
Memory Usage29.6 GB
Weights
KV Cache
Overhead
Context Length131,072 tokens
Install with Ollama:
ollama run qwen2.5-coder:32b-q6_k
Memory Usage22.5 GB
Weights
KV Cache
Overhead
Context Length131,072 tokens
Install with Ollama:
ollama run mistral-small:24b-q6_k