The ollama 0.12.6-rc0 software released this evening and with it comes experimental Vulkan API support.
The ollama software continues to be popular with enthusiasts for easily running large language models like GPT-OSS, DeepSeek-R1, Gemma 3, and of course Llama 3/4 LLMs. Ollama enjoys widespread app integration and library support while leveraging Llama.cpp for much of the heavy lifting. One long awaited feature is finally available with ollama: Vulkan API support for cases where GPU support isn’t otherwise available.
The ollama 0.12.6 release is introducing experimental Vulkan support. With this 0.12.6-rc0 test release, Vulkan support is available for users building from source. Vulkan support will eventually come to future ollama binary releases but they are currently working through various obstacles.
Vulkan support with ollama opens up usage to more GPUs from primarily AMD and Intel where ROCM or SYCL / OpenCL support isn’t otherwise available.
This ticket has been tracking Vulkan support for ollama the past year and a half and was finally closed today with the -rc0 milestone.
More details on the experimental Vulkan support for ollama via the v0.12.6-rc0 announcement.
