Best local AI models for NVIDIA 920MX

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 920MX is an entry level graphics card equipped with 2 GB of DDR3 video memory. This hardware configuration limits the size of the artificial intelligence models you can run entirely on the graphics processor. To execute a model locally without system memory slowdowns, the entire model weights must fit inside this 2 GB limit. This page lists the largest models that can run on this hardware.

The quant column indicates the quantization level used to compress the model. Quantization reduces the precision of the model weights to save memory. A lower quantization level like Q4_K_M allows larger models to fit into the 2 GB video memory but reduces output quality. A higher quantization level like Q8_0 preserves original model quality but requires more memory per parameter.

Models like Allegro 2.8B and Open-Sora Plan 2.7B represent the absolute limit of your hardware at Q4_K_M quantization. These models use exactly 2 GB of video memory. Other options like LFM2 2.6B and Playground v2.5 2.6B use 1.9 GB of video memory at Q4_K_M. If you want better output quality, you can choose smaller models with higher quantization levels. For example, SmolLM2 1.7B fits at Q6_K using 1.7 GB of video memory, while TinyLlama 1.1B fits at Q8_0 using 1.4 GB of video memory.

When a model exceeds the 2 GB video memory limit, you must use CPU offload. This technique splits the model between your graphics card and your system memory. We assume a system with 32 GB of system RAM for these cases. For example, SmolLM3 3B and Kandinsky 3.1 require 2.2 GB of video memory at Q4_K_M and 4.2 GB of system RAM. Stable Diffusion XL requires 4.1 GB at FP8 or optimized settings and 6.1 GB of system RAM. CPU offload allows you to run these larger models, but it significantly reduces processing speed.

You must also consider the memory cost of context length. Running text models with a standard 4k context window requires additional memory for the key value cache. This extra memory usage is not included in the base model weights listed above. If your video memory is already at the 2 GB limit, processing long text prompts will cause the system to slow down or fail.