Skip to content

AMD Radeon RX 9070 XT

AMD · 16GB GDDR6 · Can run 49 models

Buy Amazon
Manufacturer AMD
VRAM 16 GB
Memory Type GDDR6
Architecture RDNA 4
Stream Procs 4,096
Bandwidth 650 GB/s
TDP 300W
MSRP $549
Released Jan 23, 2025

AI Notes

The RX 9070 XT is AMD's first RDNA 4 GPU with 16GB VRAM and improved memory bandwidth. It handles 13B models comfortably and can run larger models with quantization. ROCm support for RDNA 4 is still maturing, so expect some setup effort compared to NVIDIA CUDA. Strong value at its price point for AI workloads.

Compatible Models

Model Parameters Best Quant VRAM Used Fit Est. Speed
Qwen 3 0.6B 600M Q4_K_M 2.5 GB Runs ~260 tok/s
Gemma 3 1B 1B Q8_0 2 GB Runs ~325 tok/s
Llama 3.2 1B 1B Q8_0 3 GB Runs ~217 tok/s
DeepSeek R1 1.5B 1.5B Q8_0 3 GB Runs ~217 tok/s
Gemma 2 2B 2B Q8_0 4 GB Runs ~163 tok/s
Gemma 3n E2B 2B Q4_K_M 3.3 GB Runs ~197 tok/s
Llama 3.2 3B 3B Q8_0 5 GB Runs ~130 tok/s
Phi-3 Mini 3.8B 3.8B Q8_0 5.8 GB Runs ~112 tok/s
Phi-4 Mini 3.8B 3.8B Q4_K_M 4.5 GB Runs ~144 tok/s
Gemma 3 4B 4B Q4_K_M 5 GB Runs ~130 tok/s
Gemma 3n E4B 4B Q4_K_M 4.5 GB Runs ~144 tok/s
Qwen 3 4B 4B Q4_K_M 4.5 GB Runs ~144 tok/s
DeepSeek R1 7B 7B Q8_0 9 GB Runs ~72 tok/s
Falcon 3 7B 7B Q4_K_M 6.8 GB Runs ~96 tok/s
Mistral 7B 7B Q8_0 9 GB Runs ~72 tok/s
Qwen 2.5 7B 7B Q8_0 9 GB Runs ~72 tok/s
Qwen 2.5 Coder 7B 7B Q8_0 9 GB Runs ~72 tok/s
Qwen 2.5 VL 7B 7B Q4_K_M 7 GB Runs ~93 tok/s
Cogito 8B 8B Q4_K_M 7.5 GB Runs ~87 tok/s
DeepSeek R1 8B 8B Q4_K_M 7.5 GB Runs ~87 tok/s
Llama 3.1 8B 8B Q8_0 10 GB Runs ~65 tok/s
Nemotron 3 Nano 8B 8B Q4_K_M 7.5 GB Runs ~87 tok/s
Qwen 3 8B 8B Q4_K_M 7.5 GB Runs ~87 tok/s
Gemma 2 9B 9B Q8_0 11 GB Runs ~59 tok/s
Falcon 3 10B 10B Q4_K_M 8.5 GB Runs ~76 tok/s
Llama 3.2 Vision 11B 11B Q4_K_M 8.5 GB Runs ~76 tok/s
Gemma 3 12B 12B Q4_K_M 10.5 GB Runs ~62 tok/s
Mistral Nemo 12B 12B Q4_K_M 9.5 GB Runs ~68 tok/s
DeepSeek R1 14B 14B Q4_K_M 9.9 GB Runs ~66 tok/s
Phi-4 14B 14B Q4_K_M 9.9 GB Runs ~66 tok/s
Phi-4 Reasoning 14B 14B Q4_K_M 11 GB Runs ~59 tok/s
Qwen 2.5 14B 14B Q4_K_M 9.9 GB Runs ~66 tok/s
Qwen 2.5 Coder 14B 14B Q4_K_M 12 GB Runs ~54 tok/s
Qwen 3 14B 14B Q4_K_M 12 GB Runs ~54 tok/s
Codestral 22B 22B Q4_K_M 14.7 GB Runs (tight) ~44 tok/s
StarCoder2 15B 15B Q8_0 17 GB CPU Offload ~11 tok/s
Devstral 24B 24B Q4_K_M 17 GB CPU Offload ~11 tok/s
Magistral Small 24B 24B Q4_K_M 17 GB CPU Offload ~11 tok/s
Mistral Small 3.1 24B 24B Q4_K_M 18 GB CPU Offload ~11 tok/s
Gemma 2 27B 27B Q4_K_M 17.7 GB CPU Offload ~11 tok/s
Gemma 3 27B 27B Q4_K_M 20 GB CPU Offload ~10 tok/s
Qwen 3 30B-A3B (MoE) 30B Q4_K_M 22 GB CPU Offload ~9 tok/s
Cogito 32B 32B Q4_K_M 21.5 GB CPU Offload ~9 tok/s
DeepSeek R1 32B 32B Q4_K_M 20.7 GB CPU Offload ~9 tok/s
Qwen 2.5 32B 32B Q4_K_M 20.7 GB CPU Offload ~9 tok/s
Qwen 2.5 Coder 32B 32B Q4_K_M 23 GB CPU Offload ~8 tok/s
Qwen 3 32B 32B Q4_K_M 23 GB CPU Offload ~8 tok/s
QwQ 32B 32B Q4_K_M 21.5 GB CPU Offload ~9 tok/s
Command R 35B 35B Q4_K_M 22.5 GB CPU Offload ~9 tok/s
20 model(s) are too large for this hardware.