The Radeon RX 9070 XT has 16 GB of GDDR6 and 640 GB/s of bandwidth. On paper it runs the same models as other 16 GB cards, though software setup on AMD can be less uniform than on NVIDIA.
Check your own setup with the Can My GPU Run This AI Model? calculator. The numbers below come from the same data.
RX 9070 XT specs that matter for local AI
- Memory: 16 GB GDDR6 on a 256-bit bus
- Memory bandwidth: 640 GB/s
- Board power: 304 W
VRAM decides which models fit. Bandwidth decides how fast they generate text, because every new token means reading the whole model from memory.
Which AI models fit on the RX 9070 XT?
The table assumes Q4_K_M quantization (about 4.9 bits per weight) and an 8K context. Totals include the KV cache and about 1 GB of runtime overhead. Speed is an estimate: memory bandwidth divided by model size, at 60 to 80 percent efficiency, which matches published llama.cpp results on several RTX cards.
| Model | Weights | KV cache (8K) | Total (est.) | On 16 GB | Speed (est.) |
|---|---|---|---|---|---|
| Qwen3 4B | 2.5 GB | 1.2 GB | 4.7 GB | Fits | 156 to 208 tok/s |
| Llama 3.1 8B Instruct | 4.9 GB | 1.1 GB | 7.0 GB | Fits | 78 to 104 tok/s |
| Qwen3 8B | 5.0 GB | 1.2 GB | 7.2 GB | Fits | 77 to 102 tok/s |
| Gemma 3 12B Instruct | 7.5 GB | 0.9 GB | 9.3 GB | Fits | 51 to 69 tok/s |
| Qwen3 14B | 9.0 GB | 1.3 GB | 11.4 GB | Fits | 42 to 57 tok/s |
| OpenAI gpt-oss-20b | 12.8 GB | 0.2 GB | 14.0 GB | Fits | Varies (MoE) |
| Mistral Small 3.2 24B | 14.7 GB | 1.3 GB | 17.0 GB | No | – |
| Qwen3.6 27B | 17.0 GB | 0.5 GB | 18.6 GB | No | – |
| Gemma 3 27B Instruct | 16.8 GB | 1.1 GB | 18.9 GB | No | – |
| Qwen3 32B | 20.1 GB | 2.1 GB | 23.2 GB | No | – |
| Qwen3.6 35B-A3B | 22.0 GB | 0.2 GB | 23.2 GB | No | – |
“Fits” means the estimate is at most 92 percent of VRAM. “Tight” means it is between 92 and 100 percent, so close other GPU apps. Mixture-of-Experts models (gpt-oss-20b and Qwen3.6 35B-A3B) need their full weights in memory, but they only use part of them per token, so speed is not estimated.
Best models for the RX 9070 XT
- Qwen3 14B (Q4_K_M): about 11.4 GB at 8K.
- gpt-oss-20b: about 14 GB.
- 8B and 12B models: comfortable, with Q8_0 of an 8B model at about 10.6 GB.
How context length changes things
The KV cache grows with every token of context, so a model that fits at 4K can fail at 32K.
| Model (Q4_K_M) | Context | KV cache | Total (est.) | On 16 GB |
|---|---|---|---|---|
| Llama 3.1 8B Instruct | 4K | 0.54 GB | 6.5 GB | Fits |
| Llama 3.1 8B Instruct | 8K | 1.07 GB | 7.0 GB | Fits |
| Llama 3.1 8B Instruct | 16K | 2.15 GB | 8.1 GB | Fits |
| Llama 3.1 8B Instruct | 32K | 4.29 GB | 10.2 GB | Fits |
| Qwen3 14B | 4K | 0.67 GB | 10.7 GB | Fits |
| Qwen3 14B | 8K | 1.34 GB | 11.4 GB | Fits |
| Qwen3 14B | 16K | 2.68 GB | 12.7 GB | Fits |
| Qwen3 14B | 32K | 5.37 GB | 15.4 GB | Tight |
A Q8 KV cache (an option in Ollama and llama.cpp) roughly halves the cache cost if you need more context.
What does not fit
24B to 32B models need 17 to 23 GB at Q4, which is more than 16 GB.
A good 16 GB card if you are comfortable with AMD software. It has the same memory capacity as NVIDIA’s 16 GB cards.
Verdict
What speed to expect
At 640 GB/s, an 8B Q4 model has a ceiling near 130 tokens per second. Applying the 60 to 80 percent efficiency range seen in published llama.cpp tests gives about 78 to 104 tokens per second for Llama 3.1 8B and 51 to 69 for Gemma 3 12B. We did not find a reliable measured result for this exact card, so these are estimates that could be off by 10 to 20 percent. Real speed on AMD also depends on the backend (HIP or Vulkan), so treat the upper end as optimistic.
Frequently asked questions
Can the RX 9070 XT run local AI?
Yes. llama.cpp supports AMD GPUs through HIP and Vulkan, though setup varies.
Can it run gpt-oss-20b?
Yes, with about 14 GB used.
Can it run a 27B model?
Not at Q4: it needs about 19 GB.
Sources and method
- RX 9070 XT specs: Tom’s Hardware, RDNA 4 / RX 9000 specifications (16 GB, 640 GB/s, 304 W).
- Linear-attention and sliding-window layers are handled with reduced KV cache; small fixed states in some newer models are not counted.
- Model architecture: each model’s published
config.jsonon Hugging Face (Llama 3.1 8B and Gemma 3 are read from the unsloth mirrors because the original repositories require a login). - VRAM and speed figures are estimates from our calculator, not our own benchmarks. Last updated 3 Nov 2026.
