Best local AI models for AMD R5 M435

The largest models that fit fully

The 30 largest of the 56 models that fit; every smaller model in the catalog fits too. Best quant means the highest quality compression whose weights and 4k context both sit inside the memory.

Close, but only with CPU offload

These need more than the card holds at their smallest practical quant, so part of the model runs from system memory (figures assume 32 GB of it). They work, several times slower.

How to read this

The AMD Radeon R5 M435 is an entry level laptop graphics card equipped with 2 GB of GDDR5 dedicated video memory. This hardware configuration places strict limits on the size of the AI models you can run entirely on the GPU. To run a model without system slowdowns, the model weights and the active memory space must fit completely within this 2 GB limit.

The quant column shows the quantization level used to compress each model. Quantization reduces the precision of the model weights to make them smaller. For example, a Q4_K_M quant uses approximately four bits per weight, which allows larger models like the 2.8B Allegro or the 2.7B Open-Sora Plan to fit into 2 GB of video memory. Higher quants like Q6_K or Q8_0 offer better output quality but require more memory space, limiting you to smaller models like the 1.1B TinyLlama.

When a model exceeds the 2 GB video memory limit, you must use CPU offloading. This process splits the model between your graphics card and your system RAM. If you have 32 GB of system RAM, you can run larger models like the 3.5B SDXL Turbo or the 3.3B MusicGen. However, offloading comes with a performance cost. Moving data between the system RAM and the GPU is much slower than using dedicated GDDR5 memory, which significantly reduces generation speeds.

Running text models also requires memory for the context window. The memory figures listed here represent the model weights alone. If you increase the context window to 4k tokens or higher, the active memory usage will grow. This extra memory demand can easily push a tight model over the 2 GB threshold and trigger slow system RAM offloading.

For the best performance on this hardware, you should target models that fit entirely within the 2 GB limit. Models like the 1.7B SmolLM2 at Q6_K or the 1.1B SantaCoder at Q8_0 run efficiently because they leave enough room for processing. If you need to run larger 3B models, ensure your software is configured to offload the excess layers to your system RAM.