Skip to content
Llama Builds

AMD · RDNA 4 · 2025

AMD Radeon RX 9070 XT for local AI

A solid 16GB RDNA 4 card with improving ROCm support, trailing the RTX 5070 Ti on bandwidth.

Updated · Speeds marked ~ are estimates; seemethodology

VRAM
16GB
Bandwidth
640 GB/s
Power
304W
Runs well
6 models

RX 9070 XT

Typically about $800

Check price at AmazonShop used on eBay

Can the RX 9070 XT run LLMs?

The RX 9070 XT has 16GB of GDDR6 and 640 GB/s of memory bandwidth, which decides how big a model fits and how fast it generates text. It comfortably runs models up to about 15B parameters: the largest in our catalog is Phi-4 at Q5_K_M (13GB with 8K context, ~37 tokens/sec). Our top all-round pick for it is Qwen3.8 27B (Q3_K_M, ~30 tokens/sec). Llama 3.3 70B Instruct only runs with partial CPU offload, at about 1.4 tokens/sec with CPU offload.

Recommendations

Best models for the RX 9070 XT

  1. Best overall · also coding, reasoning, vision, largest that fits

    Qwen3.8 27B

    Q3_K_M · 15GB · ~30 tokens/sec

    Fits, tight

  2. Runner-up

    Qwen3.5 9B

    Q8_0 · 12GB · ~36 tokens/sec

    Runs well

  3. Fastest good model

    Gemma 4 26B A4B

    Q3_K_M · 14GB · ~90 tokens/sec

    Fits, tight
Read: the best LLMs to run on the RX 9070 XT

Compatibility

What runs on the RX 9070 XT

Every current model in our catalog, at the best quantization that fits in 16GB with 8K tokens of context. "Partial offload" means part of the model runs from system RAM (we assume 64GB), which is much slower.

Which open-weight models run on the AMD Radeon RX 9070 XT, with the best quantization and estimated speed
ModelParametersBest quantMemorySpeedVerdict
Gemma 4 E4B8BQ8_010 GB~41 tok/sRuns well
Qwen3.5 9B9BQ8_012 GB~36 tok/sRuns well
Phi-414.7BQ5_K_M13 GB~37 tok/sRuns well
gpt-oss-20b20.9B (3.6B active)MXFP415 GB~75 tok/sFits, tight
Gemma 4 26B A4B26B (3.8B active)Q3_K_M14 GB~90 tok/sFits, tight
Qwen3.8 27B27BQ3_K_M15 GB~30 tok/sFits, tight
Gemma 4 31B31BQ4_K_M22 GB~5.4 tok/sPartial offload
Qwen3.6 35B-A3B35B (3B active)Q4_K_M24 GB~41 tok/sPartial offload
GLM-4.5-Air106B (12B active)———Won't fitWhat it needs
Llama 4 Scout109B (17B active)———Won't fitWhat it needs
gpt-oss-120b117B (5.1B active)———Won't fitWhat it needs
Qwen3.5 122B-A10B122B (10B active)———Won't fitWhat it needs
DeepSeek V4 Flash284B (13B active)———Won't fitWhat it needs
GLM-5.3-Flash320B (18B active)———Won't fitWhat it needs
GLM-5.3744B (40B active)———Won't fitWhat it needs

Performance

RX 9070 XT vs. alternatives on Qwen3.8 27B

Generation speed in tokens per second for a single user. Token generation is limited mostly by memory bandwidth, which is why bandwidth predicts these results so well.

  1. RTX 5070 Ti

    Q3_K_M · 16GB

    ~51tok/s

  2. RX 7900 XTX

    Q4_K_M · 24GB

    ~36tok/s

  3. RX 9070 XT

    Q3_K_M · 16GB

    ~30tok/s

  4. R9700

    Q6_K · 32GB

    ~18tok/s

Estimated (method)

Measured

RX 9070 XT benchmark results

ModelQuantRuntimeGenerationPromptSource
Llama 2 7B ChatQ4_0llama.cpp (Vulkan)135 tok/s5035 tok/sllama.cpp discussion #10879: Performance of llama.cpp with Vulkan · October 2026
Llama 2 7B ChatQ4_0llama.cpp (ROCm (HIP))100 tok/s5055 tok/sllama.cpp discussion #15021: Performance of llama.cpp on AMD ROCm (HIP) · October 2026

Our take

Should you buy the RX 9070 XT for AI?

The Radeon RX 9070 XT uses the Navi 48 chip with 16GB of GDDR6 on a 256-bit bus for 640 GB/s. RDNA 4 adds much faster AI matrix hardware than RDNA 3, ROCm now supports the card on Linux, and Vulkan works everywhere. It runs 14B models at 8 bit and 24B models at 4 bit, but 16GB limits anything larger. Prices rose to about $800 new in October 2026 after AMD raised RX 9000 pricing, which narrows its value advantage, though it still costs about $400 less than an RTX 5070 Ti.

What we like

  • Cheaper than NVIDIA's comparable 16GB cards
  • Much faster AI hardware than RDNA 3
  • Good Vulkan and ROCm support in llama.cpp

Watch out for

  • 640 GB/s trails the RTX 5070 Ti's 896 GB/s
  • Some tools remain CUDA-only
  • 16GB limits model size

Best for: AMD gaming PCs that also run local models · Linux users who prefer open drivers

Specifications

RX 9070 XT specs that matter for AI

Memory
16GB GDDR6
Memory bandwidth
640 GB/s
Memory bus
256-bit
Architecture
RDNA 4
Power
304W
Interface
PCIe 5.0 x16
Released
2025
Launch price
$599
Street price (new)
~$800 (2026-10)
Street price (used)
~$720 (2026-10)
Software
ROCm, Vulkan

Sources: Wikipedia, Radeon RX 9000 series

Cross-shop

Alternatives to the RX 9070 XT

NVIDIA2025
16GB

GDDR7 · 896 GB/s

RTX 5070 Ti

~$1,200

AMD2025
32GB

GDDR6 · 640 GB/s

R9700

~$1,800

AMD2022
24GB

GDDR6 · 960 GB/s

RX 7900 XTX

~$1,050 used

Frequently asked questions

How much VRAM does the RX 9070 XT have?

The AMD Radeon RX 9070 XT has 16GB of GDDR6 with 640 GB/s of memory bandwidth on a 256-bit bus.

What is the largest LLM the RX 9070 XT can run?

Fully in GPU memory, the largest model in our catalog that runs comfortably is Phi-4 (14.7B) at Q5_K_M, using about 13GB with 8K context at ~37 tokens/sec.

Can the RX 9070 XT run Llama 3.3 70B?

Only with partial CPU offload. At Q4_K_M the model needs about 47GB with 8K context, so part of it spills into system RAM and speed drops to about 1.4 tokens/sec with CPU offload. For usable 70B speeds you want 48GB of VRAM or more (for example two 24GB cards).

What is the best LLM to run on the RX 9070 XT?

Our top pick is Qwen3.8 27B at Q3_K_M, which fits in 16GB with room for 8K of context and runs at ~30 tokens/sec. See our full list of the best models for the RX 9070 XT for coding, reasoning, and vision picks.

Is the RX 9070 XT good for local AI in 2026?

A solid 16GB RDNA 4 card with improving ROCm support, trailing the RTX 5070 Ti on bandwidth.