Best local AI models for NVIDIA MX110

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 MX110 is an entry level laptop graphics card equipped with 2 GB GDDR5 video memory. This memory size is the absolute limit for what can run directly on the graphics hardware. To run artificial intelligence models locally on this card, you must select small models and use compressed versions. Trying to load a model that exceeds this limit will cause execution to fail or slow down significantly.

The quantization column indicates the compression level applied to each model. Quantization reduces the precision of the model weights to save space. For example, a Q4_K_M quantization represents a four bit format that balances model size and output quality. Higher quantizations like Q6_K or Q8_0 offer better accuracy but require more video memory. For the NVIDIA MX110, you must choose the right quantization to fit the 2 GB limit.

Several highly optimized models can fit entirely within the 2 GB video memory of the NVIDIA MX110. The Allegro 2.8B model fits at Q4_K_M quantization while using exactly 2 GB. The Open-Sora Plan 2.7B model also fits at Q4_K_M quantization with 2 GB used. You can also run the LFM2 2.6B model or the Playground v2.5 2.6B model at Q4_K_M quantization, which both use 1.9 GB of video memory.

Other lightweight options are available for different tasks. The Stable Diffusion 3.5 Medium 2.5B model and the Canary 2.5B model both use 1.8 GB of video memory at Q4_K_M quantization. For text to speech, the Parler-TTS 2.2B model fits at Q5_K_M quantization using 1.9 GB. Vision tasks can use the SmolVLM 2B model at Q6_K quantization, which uses exactly 2 GB of video memory.

When a model is slightly too large for the video memory, you can use CPU offload. This technique splits the model between the graphics card and your system memory. For CPU offload, we assume your computer has 32 GB of system RAM. For example, the SmolLM3 3B model needs 2.2 GB of video memory at Q4_K_M quantization and requires 4.2 GB of system RAM. The MusicGen 3.3B model needs 2.4 GB of video memory at Q4_K_M quantization and 4.4 GB of system RAM.

Using CPU offload comes with a performance cost. Moving data between the system RAM and the graphics card is much slower than keeping everything inside the video memory. Also, you must consider the context window. Running models with a standard 4k context window increases memory usage during active generation. If you generate long text or process large inputs, the memory usage will exceed the base figures listed here.