Best local AI models for AMD R7 M260

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 R7 M260 is an entry level mobile graphics card equipped with 2 GB of DDR3 video memory. This dedicated memory pool is the primary constraint when running artificial intelligence models locally. Because DDR3 memory has lower bandwidth than modern GDDR standards, keeping the entire model inside the video memory is crucial for maintaining acceptable processing speeds.

To fit inside the 2 GB limit, models must undergo quantization. The quant column indicates the compression level applied to the model weights. For example, the Allegro 2.8B model uses a Q4_K_M quantization to fit exactly 2 GB of video memory. Smaller models like SmolLM2 1.7B can run at a higher quality Q6_K quantization while using 1.7 GB of memory. TinyLlama 1.1B runs at a high fidelity Q8_0 quantization using only 1.4 GB of video memory.

When a model size exceeds the local video memory, you must use CPU offload. This technique splits the workload between your graphics card and your system RAM. Assuming your computer has 32 GB of system RAM, you can run larger models like the 3B SmolLM3 or the 3.5B SDXL Turbo. However, offloading introduces a speed penalty because data must constantly travel over the system bus.

For instance, running the 3B Kandinsky 3.1 or Orpheus TTS requires 2.2 GB of video memory at Q4_K_M and an additional 4.2 GB of system RAM. Larger generation models like Stable Diffusion XL require 4.1 GB at FP8 or optimized settings, which demands 6.1 GB of system RAM. While offloading makes these larger models run, the processing time will be significantly longer than models that fit entirely within the 2 GB video memory.

Users must also consider the memory cost of context length. Running text models with a standard 4k context window increases the active memory footprint during generation. If your model already uses 1.9 GB or 2 GB of video memory, like the Open-Sora Plan 2.7B or the Kimi K3 DSpark 2.2B, the additional memory required for context processing will force the system to spill over into system RAM, slowing down your generation speeds.