Best local AI models for NVIDIA 920M

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 NVIDIA 920M is an entry level mobile graphics card equipped with 2 GB of DDR3 video memory. This limited memory capacity dictates which artificial intelligence models you can run locally. To load a model entirely on the graphics card, the total memory footprint of the model must remain under the 2 GB hardware limit. Running models directly in this dedicated video memory ensures the fastest possible processing speeds on this specific hardware.

The quantization column indicates the compression level applied to each model. Quantization reduces the precision of model weights to save space. For example, Allegro 2.8B and Open-Sora Plan 2.7B fit within 2 GB of video memory when compressed to the Q4_K_M quantization level. Smaller models like SmolLM2 1.7B or Qwen3 1.7B can run at a higher Q6_K quantization. Ultra small models such as TinyLlama 1.1B or SantaCoder 1.1B can run at the high quality Q8_0 quantization while using only 1.4 GB of video memory.

When a model exceeds the 2 GB video memory limit, you must use CPU offloading. This technique splits the model weights between your graphics card and your system RAM. For instance, running SmolLM3 3B or Kandinsky 3.1 requires 2.2 GB of video memory at Q4_K_M quantization and an additional 4.2 GB of system RAM. Larger options like SDXL Turbo 3.5B require 2.6 GB of video memory and 4.6 GB of system RAM. CPU offloading allows you to run these larger models but it introduces a significant performance cost because system RAM is much slower than video memory.

System RAM capacity is critical for offloading scenarios. The calculations on this page assume your computer has 32 GB of system RAM installed. If your system has less RAM, you may experience severe slowdowns or system instability when offloading. For heavy models like Stable Diffusion XL which requires 4.1 GB at FP8 or optimized settings along with 6.1 GB of system RAM, having sufficient system memory is the only way to prevent crashes.

You must also consider the context window when running local text models. The memory usage figures listed here represent the base model size. Generating long responses or inputting large prompts increases memory consumption. If you use a standard 4k context window, the active memory will expand beyond the base figures. You may need to select a smaller model or a lower quantization level to prevent your system from running out of memory during long conversations.