Best local AI models for AMD HD 7730M

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 HD 7730M is a mobile graphics processor equipped with 2 GB of DDR3 video memory. This dedicated memory capacity dictates the maximum size of the artificial intelligence models you can run entirely on the graphics card. To execute models locally without system memory bottlenecks, the model files and their active working memory must fit within this 2 GB limit.

Quantization is a compression method that reduces the precision of model weights to save space. The best quant column indicates the optimal balance of size and quality for this hardware. For example, the 2.8B Allegro and 2.7B Open-Sora Plan models require a Q4_K_M quantization to fit exactly 2 GB of used video memory. Models like the 2.2B Parler-TTS can run at a higher quality Q5_K_M quantization using 1.9 GB of video memory.

Smaller models can leverage even higher precision quantizations on this hardware. The 2B Stable Diffusion 3 Medium and 2B SmolVLM use a Q6_K quantization to occupy exactly 2 GB of video memory. Ultra compact models such as the 1.1B TinyLlama and 1.1B SantaCoder can run at the high quality Q8_0 quantization level while using only 1.4 GB of video memory.

When a model exceeds the 2 GB video memory limit, you must offload the remaining parameters to your system RAM. This offload process allows you to run larger models but introduces a performance cost because DDR3 video memory and system RAM must constantly exchange data. For instance, running the 3.5B SDXL Turbo at Q4_K_M requires 2.6 GB of video memory and 4.6 GB of system RAM.

Other offload examples include the 3B SmolLM3 and 3B Kandinsky 3.1 which require 2.2 GB of video memory and 4.2 GB of system RAM at Q4_K_M. The 3.3B MusicGen requires 2.4 GB of video memory and 4.4 GB of system RAM. Running the 3.417B Stable Diffusion XL at FP8 requires 4.1 GB of video memory and 6.1 GB of system RAM.

You must also consider the memory cost of context length when running language models. The memory figures listed here are calculated for minimal context. If you increase the context window to 4k tokens, the active memory usage will rise. This extra memory demand can exceed the 2 GB limit of the AMD HD 7730M and force system RAM offloading even for smaller models.