AMD Ryzen AI and Radeon to Experience the DeepSeek-R1 Distilled ‘Reasoning’ Models.
A new type of large language models (LLMs) called reasoning models uses chain-of-thought (CoT) reasoning to solve extremely complicated tasks, but at the expense of responding more slowly. Recently published, the DeepSeek-R1 is a frontier “reasoning” model that has been condensed into incredibly powerful smaller models. These DeepSeek R1 distillation models are now available through LM Studio and are very simple to deploy on AMD Ryzen AI CPUs and Radeon graphics cards.
Real-time demonstration of the DeepSeek-R1 Qwen 1.5 Q4 K M model operating on an AMD Ryzen HX 370 series processor.
Expanding the “thinking” window prior to the model’s final response reveals that reasoning models include a “thinking” phase before the final output. CoT LLMs conduct in-depth reasoning prior to responding, in contrast to traditional LLMs that provide a single response. The user may see the assumptions and self-reflection the LLM makes, which enhances the model’s reasoning and analytical power, but at the expense of a noticeably longer time-to-first- (final output) token.
Before providing a final answer, a reasoning model may analyze the issue for thousands of tokens (you can see this line of reasoning!). This enables the model to approach a complex problem from multiple perspectives before choosing a solution, making it very good at challenging math and science-related problem-solving activities. Each of these models will provide cutting-edge reasoning capabilities on your AMD Radeon graphics cards or Ryzen AI processor, depending on your AMD hardware.
How to use AMD Ryzen AI and Radeon Graphics Cards with DeepSeek-R1 Distilled “Reasoning” Models
Get started with DeepSeek R1 distillations in a matter of minutes by following these easy steps (depending on download speed).
Step 1: Verify that you are running Adrenalin driver 25.1.1 or above.
Step 2: Go to lmstudio.ai/ryzenai and download LM Studio 0.3.8 or later.
Step 3: Install LM Studio in step three, avoiding the onboarding page.
Step 4: Select the tab labelled “discover.”
Step 5: Select your DeepSeek-R1 Distil in step five. While larger distillers will have higher reasoning capability, smaller distillers, such as the Qwen 1.5B, offer lightning-fast performance (and are the suggested starting point). They are all quite competent. The maximum recommended DeepSeek R1 Distil size is shown in the table below:
| Processor | DeepSeek R1 Distill (Max Supported) |
| AMD Ryzen AI Max+ 395 32GB, 64 GB and 128 GB | DeepSeek-R1-Distill-Llama-70B (64GB and 128GB only) DeepSeek-R1-Distill-Qwen-32B |
| AMD Ryzen AI HX 370 and 365 24GB and 32 GB | DeepSeek-R1-Distill-Qwen-14B |
| AMD Ryzen 8040 and Ryzen 7040 32 GB | DeepSeek-R1-Distill-Llama-14B |
- All distils should be conducted in Q4 K M quantisation, according to AMD.
- 24GB of custom graphics memory is selected.
- Variable Graphics Memory must be set to High.
| Graphics Card | DeepSeek R1 Distill (Max Supported) |
| AMD Radeon RX 7900 XTX | DeepSeek-R1-Distill-Qwen-32B |
| AMD Radeon RX 7900 XT | DeepSeek-R1-Distill-Qwen-14B |
| AMD Radeon RX 7900 GRE | DeepSeek-R1-Distill-Qwen-14B |
| AMD Radeon RX 7800 XT | DeepSeek-R1-Distill-Qwen-14B |
| AMD Radeon RX 7700 XT | DeepSeek-R1-Distill-Qwen-14B |
| AMD Radeon RX 7600 XT | DeepSeek-R1-Distill-Qwen-14B |
| AMD Radeon RX 7600 | DeepSeek-R1-Distill-Llama-8B |
- All distils should be conducted in Q4 K M quantization, according to AMD.
- lists the highest distillation that can be done without a substantial GPU offload.
Step 6: Click “Download” after making sure the “Q4 K M” quantization is chosen on the right-hand side.
Step 7: Return to the chat tab after downloading, choose the DeepSeek-R1 distil from the drop-down box, and ensure that “manually select parameters” is selected.
Step 8: Move the slider to the maximum in the GPU offload layers.
Step 9: Click “model load” in step nine.
Step 10: Engage with a reasoning model that is fully operational on AMD hardware in your area!
Conclusions
GD-97: Unless otherwise indicated, AMD is not liable for the content of linked websites and makes no implicit endorsements.
The AMD Radeon graphics engine, a specialized AI engine, and Ryzen CPU cores that provide AI capabilities make up GD-220e-Ryzen AI. Some AI functions might not be optimized for Ryzen AI CPUs yet, therefore OEM and ISV enablement is necessary. Ryzen AI can be used with:
- AMD Ryzen 7040 and 8040 Series processors and Ryzen PRO 7040/8040 Series processors, with the exception of Ryzen 5 7540U, Ryzen 5 8540U, Ryzen 3 7440U, and Ryzen 3 8440U processors.
- AMD Ryzen AI 300 Series processors and AMD Ryzen AI PRO 300 Series processors.
- All AMD Ryzen 8000G Series desktop processors, with the exception of Ryzen 5 8500G/GE and Ryzen 3 8300G/GE.
- AMD Ryzen 200 Series processors and Ryzen PRO 200 Series processors, with the exception of Ryzen 5 220 and Ryzen 3 210.
- AMD Ryzen AI Max Series processors and Ryzen AI PRO Max Series processors. Before making a purchase, please confirm feature availability with the maker of your system.
