Can You Use OpenVINO on LM Studio?

No, LM Studio does not currently expose llama.cpp’s OpenVINO backend as a selectable runtime. Installing OpenVINO separately does not add an OpenVINO option to LM Studio.

That does not mean OpenVINO and GGUF are incompatible. OpenVINO is an option as an acceleration backend for llama.cpp, and the current llama.cpp OpenVINO backend is explicitly compatible with the some of the existing GGUF model ecosystem.

The short answer

If you want to run a GGUF model in LM Studio, use one of the llama.cpp runtimes that LM Studio provides. On Intel Arc, that means the Vulkan runtime for GPU acceleration.

If you specifically want llama.cpp + OpenVINO, run llama.cpp with its OpenVINO backend outside LM Studio. Installing OpenVINO does not make LM Studio use that backend.

What LM Studio is doing

LM Studio uses llama.cpp to run GGUF models. The acceleration backend is a separate choice from the model file:

Choice	What it means
CPU	llama.cpp running model computation on the CPU
Vulkan	llama.cpp using its Vulkan backend, including Intel Arc GPUs
OpenVINO	llama.cpp’s OpenVINO backend, not exposed by LM Studio

The backend is not determined simply by which GPU is installed. You can select and use both CUDA and Vulkan on an Nvidia GPU for example. LM Studio provides and manages its own prebuilt runtimes.

Why can’t I select OpenVINO in LM Studio?

Because LM Studio does not currently provide llama.cpp’s OpenVINO backend as one of its available runtimes.

The llama.cpp OpenVINO backend is still actively under development. Its current documentation lists incomplete accuracy validation, performance and memory work, broader quantization coverage, additional model architectures, and other limitations.

So installing OpenVINO separately will not make a new OpenVINO option appear in LM Studio. There is no Python package, environment variable, or LM Studio setting that converts LM Studio’s existing Vulkan runtime into the OpenVINO runtime.

If you specifically want llama.cpp + OpenVINO

You can run llama.cpp directly with an OpenVINO-enabled build. The official llama.cpp documentation currently provides an OpenVINO backend and examples for running GGUF files with it.

For example, the OpenVINO backend can select an Intel GPU with:

export GGML_OPENVINO_DEVICE=GPU

On a system where OpenVINO exposes multiple GPUs, GPU.0 or GPU.1 can be selected explicitly.

This is a standalone llama.cpp setup. It is not a way to add OpenVINO support to LM Studio.

What should you use on Intel Arc?

If your goal is simply to run GGUF models on an Intel Arc GPU inside LM Studio, use Vulkan. You download the GGUF model, install/select the Vulkan-enabled llama.cpp runtime in LM Studio, and run the model.

If your goal is specifically to test or use llama.cpp’s OpenVINO backend, use llama.cpp directly with an OpenVINO-enabled build or the official OpenVINO-enabled container images. The llama.cpp project currently publishes OpenVINO-enabled Docker images as well.

If your goal is to use OpenVINO itself rather than its llama.cpp backend, use an application or API built around OpenVINO. That is a separate workflow from LM Studio.

What about CPU offload?

CPU offload is unrelated to OpenVINO.

With llama.cpp, a model that does not fit entirely in the GPU can use both the GPU and CPU. Some layers can be offloaded to the GPU while other layers remain on the CPU. You can use this on LM studio by selecting CPU offload.

On an integrated Intel GPU, system RAM is shared with the GPU, so its memory situation is different from a discrete Arc card with dedicated VRAM.

The short version

OpenVINO is a llama.cpp acceleration backend, and it can run GGUF models. LM Studio simply does not expose that backend.

For Intel Arc in LM Studio, use Vulkan. You do not need to build llama.cpp yourself for that.

If you specifically want OpenVINO, use standalone llama.cpp with its OpenVINO backend, or use an application built directly around OpenVINO.

For the broader Intel Arc backend comparison, see How to Run Local LLMs on Intel Arc.