# Intel IPEX-LLM Ollama

Intel IPEX-LLM Ollama is a containerized Ollama build with Intel IPEX-LLM acceleration that lets AI enthusiasts run large language models (LLMs) quickly on Intel GPUs, NPUs, and CPUs, though the upstream IPEX-LLM. It is open source and self-hosted, installable on Unraid.

Canonical page: https://ourlocalhost.com/app/2930/intel-ipex-llm-ollama

## Description

Intel IPEX-LLM Ollama is a containerized Ollama build with Intel IPEX-LLM acceleration that lets AI enthusiasts run large language models (LLMs) quickly on Intel GPUs, NPUs, and CPUs, though the upstream IPEX-LLM project is archived and carries known security issues.

## Platform availability

Available on 1 of 10 platforms: Unraid.

## Health score

28/100 (Low), based on 4 of 4 factors.
- Maintained: 0 (GitHub repository is archived)
- Popular: 79 (8,842 GitHub stars)
- Easy to install: 35 (1 install platform, install notes, Docker image)
- Light to run: 10 (8 GB minimum RAM)

## Links

- GitHub: https://github.com/intel/ipex-llm
- Unraid: https://ca.unraid.net/apps/intel-ipex-llm-ollama-09iewo91p6y8i5?q=Intel-IPEX-LLM-Ollama

## Common questions

### What should I do if IPEX-LLM Ollama does not work on a PC with an AMD 8700F CPU?

One user reported that it works after reinstalling OneAPI and IPEX-LLM Ollama while following Intel's graphics installation guidance.

### Can model behavior change when migrating from BigDL-LLM to IPEX-LLM?

Yes, a user found that a Vicuna 13B model stopped working after migrating, while a Phi-2-super model ran at about 7-8 tokens per second.

### What trade-offs does consolidating inference into one Intel IPEX-LLM Ollama container involve?

It collapses two inference engines into a single container, which reduces moving parts and adds native FIM support, but drops the OpenVINO stack.

---
Source: Our Local Host (https://ourlocalhost.com/app/2930/intel-ipex-llm-ollama), the self-hosted and open-source app directory.
