LLM for Unity
Run local LLM characters inside Unity
- Best for
- Unity developers who want AI characters that run offline on the player's device at no per-use cost
- Makes
- Local LLM-driven characters with retrieval (RAG) and grammar-limited output
- Made by
- undream ai
- Where you use it
- Unity Editor
- Engines
- Unity
- Pricing
- Free and open source (Apache-2.0) for personal and commercial use
- Free plan
- Yes
- Open source
- Yes (Apache-2.0)
- Export
- Models and code ship inside your own Unity build; no service account is needed
- Status
- Available
LLM for Unity, or LLMUnity, is a free, open-source package from undream ai that runs large language models inside Unity games. Its backend, LlamaLib, is built on llama.cpp, so it loads models in the GGUF format and runs them on the player's CPU or GPU (NVIDIA, AMD or Apple Metal) without an internet connection, according to the project's README. It can also connect to a remote server.
Setup happens in the Inspector: add an LLM component, download a default model or load your own GGUF file, then give each character an LLMAgent component with a system prompt. The package includes a retrieval (RAG) system for searching your own lore, and you can restrict replies with a grammar, which is also how it handles function calling: the grammar limits output to function names you allow.
The README says it's tested on Unity 2021 LTS through Unity 6 and runs on PC, mobile and VR, with iOS and Android build notes. It lists more than 20 projects that use it, including Steam games such as Verbal Verdict and Case Closed. The repository was created in December 2023, is licensed under Apache-2.0 and had 1,736 stars on October 29, 2026; its latest release was v3.0.3 on March 8, 2026.
Local models are small, and they add size and memory demands to your game. AIGamer hasn't tested it; see our AI NPCs guide. For Godot, NobodyWho fills a similar role.
Plans and pricing
| Plan | Price | What you get |
|---|---|---|
| Open source | Free | Apache-2.0; on GitHub and the Unity Asset Store |
From LLM for Unity's own pricing page, checked 29 October 2026. Prices change; check before you buy.
Limitation: On mobile, the README says models of up to 1 to 2 billion parameters are the practical limit, and its default model downloads run to gigabytes.
FAQ
Is LLMUnity free?
Yes. It's open source under the Apache-2.0 license and, per its README, free for personal and commercial use.
Does LLMUnity work offline?
Yes. It runs llama.cpp models on the player's machine, so no internet connection is needed. A remote server setup is optional.
Which Unity versions does LLMUnity support?
The README says it's tested on Unity 2021 LTS, 2022 LTS, 2023 and Unity 6.
Can LLMUnity run on phones?
Yes, on iOS and Android, but the README recommends models of up to 1 to 2 billion parameters on mobile hardware.
Sources
Compiled by AIGamer from LLM for Unity's own pages and the sources below, checked 29 October 2026. We haven't tested LLM for Unity hands-on.
- undream ai: LLMUnity (GitHub README)
- GitHub: undreamai/LLMUnity repository data
Image: undream ai (LLMUnity logo)
Written in our own words; quotes are marked and credited. How we report