GPUs & Macs › AMD Radeon AI PRO R9700
AMD Radeon AI PRO R9700: local LLM speed
Llama 2 7B (Q4_0) generates 153 tokens/s on the AMD Radeon AI PRO R9700 (Vulkan), and processes prompts at 5,908 tokens/s.
Measured results
| Model | Generate tok/s | Prompt tok/s | Quant | Backend | Test | Source |
|---|---|---|---|---|---|---|
| Llama 2 7B | 153 | 5,908 | Q4_0 | Vulkan FA on | tg128 | llama.cpp #10879 details |
| Llama 2 7B | 146 | 5,610 | Q4_0 | Vulkan FA off | tg128 | llama.cpp #10879 details |
| Llama 2 7B | 98.0 | 4,773 | Q4_0 | ROCm FA on | tg128 | llama.cpp #15021 details |
| Llama 2 7B | 93.8 | 4,444 | Q4_0 | ROCm FA off | tg128 | llama.cpp #15021 details |
| Qwen2.5 14B Instruct | 4.3 | 381 | Q4_K_M | llamafile | LocalScore avg | LocalScore details |
What fits on the AMD Radeon AI PRO R9700
Estimates for this chip are scaled from a measured run: Llama 2 7B Q4_0 at 153 tok/s (Vulkan, source). Fit assumes the whole model in GPU memory with ~8K context.
| Model | Best quant that fits | File | Expected tok/s |
|---|---|---|---|
| Qwen3.8-27B | Q8_0 | 29.0 GB | 20.1 estimate |
| gemma-4-12B-it | Q8_0 | 12.7 GB | 46.1 estimate |
| Qwen3.5-9B | Q8_0 | 9.5 GB | 61.3 estimate |
| Qwen3.5-4B | Q8_0 | 4.5 GB | 130 estimate |
| Qwen3.6-35B-A3B | UD-Q6_K | 29.3 GB | 116 estimate |
| gemma-4-26B-A4B-it | Q8_0 | 26.9 GB | 70.6 estimate |
| MiMo-V2.6-Distill-Qwen-9B | Q8_0 | 9.5 GB | 61.1 estimate |
| gemma-4-E4B-it | Q8_0 | 8.2 GB | 71.2 estimate |
| Muse-Glimmer-30B | Q8_0 | 29.6 GB | 19.7 estimate |
| Qwen3-4B | Q8_0 | 4.3 GB | 136 estimate |
| gemma-4-31B-it | Q6_K | 25.2 GB | 23.2 estimate |
| gemma-4-26B-A4B-it-qat-q4_0-unquantized | UD-Q4_K_XL | 14.2 GB | 133 estimate |
| gemma-4-12B-it-qat-q4_0-unquantized | UD-Q4_K_XL | 6.7 GB | 86.9 estimate |
| gemma-3-1b-it | Q8_0 | 1.1 GB | 486 estimate |
Similar hardware
AMD Radeon Pro V620 · AMD Radeon Pro VII · AMD Radeon Pro W7900