Best local AI models for Intel Arc A770

The largest models that fit fully

The 30 largest of the 155 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 Intel Arc A770 graphics card features 16 GB of GDDR6 dedicated memory. This memory capacity determines which artificial intelligence models you can run entirely on your hardware. When a model fits completely inside the 16 GB video memory, it runs at maximum speed because the graphics processor has direct high speed access to all parameters.

The quantization column shows the optimal format for each model on this hardware. Quantization compresses the weight files of a model to save space. For example, the 21B gpt-oss-20b and Reka Flash 3 models fit in 15.4 GB using a Q4_K_M quantization. Models like DeepSeek-Coder-V2 16B / 236B and Kimi-VL A3B use a Q6_K quantization which requires 15.7 GB of memory. Smaller models like Gemma 3 12B and Mistral NeMo 12B can run at a higher Q8_0 quantization using 15.3 GB of memory.

Running models at their limit requires careful management of your hardware resources. The memory numbers listed represent the static size of the model weights in video RAM. When you generate text, the system needs additional memory to store the conversation history. Running a model very close to the 16 GB limit leaves less room for this context window, which may limit your maximum prompt length.

If you want to run larger models, you can offload some layers to your system memory. This approach assumes your computer has at least 32 GB of system RAM. Offloading allows you to run larger files, but it reduces generation speed because data must travel over the system bus between the system RAM and the graphics card.

With CPU offloading, you can run the 22B Solar Pro or Codestral 22B models which need 16.1 GB at Q4_K_M quantization and 18.1 GB of system RAM. You can also run the 24B Mistral Small 3.2, Magistral Small, or Devstral Small 1.1 models which require 17.6 GB at Q4_K_M quantization and 19.6 GB of system RAM. The largest compatible options include the Gemma 3 27B model which requires 19.8 GB at Q4_K_M quantization and 21.8 GB of system RAM.