The AI Bench
16 GB · 640 GB/s

What can the AMD Radeon RX 9070 XT run?

AMD’s fastest current 16 GB card. Good speed; software takes a little more care.

Best models for each task

Our top three for this machine. Chat and image tasks assume an 8K-token conversation; coding, documents and agents assume 32K.

Chat & writing15 of 25 run

  1. Gemma 4 12BRuns wellQ8_0 · needs 14.1 GB at 8K · 30–45tok/sSmooth
  2. Qwen3.8 27BRuns wellUD-IQ3_S · needs 14.5 GB at 8K · 30–45tok/sSmooth
  3. MiniCPM5 2BRuns greatQ8_0 · needs 4.0 GB at 8K · 110–180tok/sFast

Coding12 of 26 run

  1. Qwen3.5 4BRuns greatQ8_0 · needs 7.2 GB at 32K · 65–100tok/sFast
  2. Gemma 4 12BRuns greatQ4_K_M · needs 9.2 GB at 32K · 45–75tok/sFast
  3. Ornith 1.5 9BRuns wellQ8_0 · needs 11.9 GB at 32K · 35–55tok/sFast

Long documents9 of 20 run

  1. Qwen3.5 9BRuns wellQ8_0 · needs 12.5 GB at 32K · 35–55tok/sFast
  2. MiniCPM5 2BRuns greatQ8_0 · needs 5.1 GB at 32K · 85–140tok/sFast
  3. Gemma 4 12BRuns greatQ4_K_M · needs 9.2 GB at 32K · 45–75tok/sFast

Agents & tools12 of 26 run

  1. Qwen3.5 4BRuns greatQ8_0 · needs 7.2 GB at 32K · 65–100tok/sFast
  2. Ornith 1.5 9BRuns wellQ8_0 · needs 11.9 GB at 32K · 35–55tok/sFast
  3. MiMo-V2.6 9BRuns wellQ8_0 · needs 12.5 GB at 32K · 35–55tok/sFast

Images in6 of 13 run

  1. Qwen3.8 27BRuns wellUD-IQ3_S · needs 14.5 GB at 8K · 30–45tok/sSmooth
  2. Gemma 4 12BRuns wellQ8_0 · needs 14.1 GB at 8K · 30–45tok/sSmooth
  3. Qwen3.5 9BRuns greatQ8_0 · needs 11.7 GB at 8K · 40–60tok/sFast

Make images6 of 6 run

  1. Stable Diffusion XL Base 1.0Runs greatFP16 · needs 7.9 GB
  2. FLUX.2 [klein] 4BRuns wellBF16 · needs 14.0 GB
  3. Z-Image-TurboRuns great4-bit GGUF · needs 5.0 GB

Speak text7 of 7 run

  1. Qwen3-TTS 1.7BRuns greatBF16 · needs 5.5 GB
  2. Kokoro-82MRuns greatFP32 · needs 1.4 GB
  3. Chatterbox Multilingual V3Runs greatFull · needs 4.2 GB

Transcribe6 of 7 run

  1. Whisper large-v3-turbo (via faster-whisper)Runs greatFP16 · needs 2.6 GB
  2. Nemotron 3.5 ASR Streaming 0.6BRuns greatFP32 · needs 3.5 GB
  3. Qwen3-ASR 1.7BRuns greatBF16 · needs 5.7 GB

Biggest model it runs well

Qwen3.8 27B (UD-IQ3_S, 12.0 GB file) at about 30–45tok/sSmooth.

Should you buy it?

Buy it if

You prefer AMD and use LM Studio or llama.cpp with Vulkan.

Skip it if

You want every AI tool to just work — NVIDIA is still easier.

Watch out for

  • Read our AMD guide before buying.