LocalAI/docs/content/faq.md

+++ disableToc = false title = "FAQ" weight = 24 icon = "quiz" url = "/faq/" +++

Frequently asked questions

Here are answers to some of the most common questions.

Do I need to install all the backends?

No. You install only the backends your models use. LocalAI's core is a single binary (or container) that provides the OpenAI-compatible API, request routing, the web UI, and agents. Each inference backend (llama.cpp, vLLM, whisper.cpp, stable-diffusion, MLX, and others) is a separate artifact, installed only when a model needs it.

In practice:

• You install one backend, not all of them. Run a model with local-ai run <model> and the matching backend is pulled automatically; nothing else is downloaded.
• Each backend is purpose-built for its engine. LocalAI builds a dedicated gRPC backend around each engine, so every one stays independently optimized without a single binary trying to support every model architecture at once.
• You manage backends individually with local-ai backends list/install/uninstall or from the web UI.

The catalog's breadth is optionality: you only ever run what your models use.

Can I bring my own model or backend?

Yes. You can load any compatible model, not just the ones in the gallery. And because every backend talks to the core over a simple gRPC interface, you can write your own backend in any language and plug it in, exactly how the built-in backends work. Nothing about the core is closed off, which gives you the flexibility to run precisely the stack you want.

How do I get models?

Most gguf-based models should work, but newer models may require additions to the API. If a model doesn't work, please feel free to open up issues. However, be cautious about downloading models from the internet and directly onto your machine, as there may be security vulnerabilities in lama.cpp or ggml that could be maliciously exploited. Some models can be found on Hugging Face: https://huggingface.co/models?search=gguf, or models from gpt4all are compatible too: https://github.com/nomic-ai/gpt4all.

Where are models stored?

LocalAI stores downloaded models in the following locations by default:

• Command line: ./models (relative to current working directory)
• Docker: /models (inside the container, typically mounted to ./models on host)
• Launcher application: ~/.localai/models (in your home directory)

You can customize the model storage location using the LOCALAI_MODELS_PATH environment variable or --models-path command line flag. This is useful if you want to store models outside your home directory for backup purposes or to avoid filling up your home directory with large model files.

How much storage space do models require?

Model sizes vary significantly depending on the model and quantization level:

• Small models (1-3B parameters): 1-3 GB
• Medium models (7-13B parameters): 4-8 GB
• Large models (30B+ parameters): 15-30+ GB

Quantization levels (smaller files, slightly reduced quality):

• Q4_K_M: ~75% of original size
• Q4_K_S: ~60% of original size
• Q2_K: ~50% of original size

Storage recommendations:

• Ensure you have at least 2-3x the model size available for downloads and temporary files
• Use SSD storage for better performance
• Consider the model size relative to your system RAM - models larger than your RAM may not run efficiently

Benchmarking LocalAI and llama.cpp shows different results!

LocalAI applies a set of defaults when loading models with the llama.cpp backend, one of these is mirostat sampling - while it achieves better results, it slows down the inference. You can disable this by setting mirostat: 0 in the model config file. See also the advanced section ({{%relref "advanced/advanced-usage" %}}) for more information and this issue.

What's the difference with Serge, or XXX?

LocalAI is a multi-model solution that doesn't focus on a specific model type (e.g., llama.cpp or alpaca.cpp), and it handles all of these internally for faster inference, easy to set up locally and deploy to Kubernetes.

Everything is slow, how is it possible?

There are few situation why this could occur. Some tips are:

• Don't use HDD to store your models. Prefer SSD over HDD. In case you are stuck with HDD, disable mmap in the model config file so it loads everything in memory.
• Watch out CPU overbooking. Ideally the --threads should match the number of physical cores. For instance if your CPU has 4 cores, you would ideally allocate <= 4 threads to a model.
• Run LocalAI with DEBUG=true. This gives more information, including stats on the token inference speed.
• Check that you are actually getting an output: run a simple curl request with "stream": true to see how fast the model is responding.

Can I use it with a Discord bot, or XXX?

Yes! If the client uses OpenAI and supports setting a different base URL to send requests to, you can use the LocalAI endpoint. This allows to use this with every application that was supposed to work with OpenAI, but without changing the application!

Can this leverage GPUs?

There is GPU support, see {{%relref "features/GPU-acceleration" %}}.

Where is the webUI?

There is the availability of localai-webui and chatbot-ui in the examples section and can be setup as per the instructions. However as LocalAI is an API you can already plug it into existing projects that provides are UI interfaces to OpenAI's APIs. There are several already on Github, and should be compatible with LocalAI already (as it mimics the OpenAI API)

Does it work with AutoGPT?

Yes, see the examples!

How can I troubleshoot when something is wrong?

Enable the debug mode by setting DEBUG=true in the environment variables. This will give you more information on what's going on. You can also specify --debug in the command line.

I'm getting 'invalid pitch' error when running with CUDA, what's wrong?

This typically happens when your prompt exceeds the context size. Try to reduce the prompt size, or increase the context size.

I'm getting a 'SIGILL' error, what's wrong?

Your CPU probably does not have support for certain instructions that are compiled by default in the pre-built binaries. If you are running in a container, try setting REBUILD=true and disable the CPU instructions that are not compatible with your CPU. For instance: CMAKE_ARGS="-DGGML_F16C=OFF -DGGML_AVX512=OFF -DGGML_AVX2=OFF -DGGML_FMA=OFF" make build