> For the complete documentation index, see [llms.txt](https://yvechat.gitbook.io/wp/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://yvechat.gitbook.io/wp/getting-started/system-requirements.md).
# System Requirements
Requirements depend on the models you intend to run. A machine that comfortably runs a small model may struggle with a large one.
The short checklist before you begin
***
### Baseline
| Component | Requirement |
| ---------------- | ------------------------------------------------------------------------------------------------------------------------ |
| Operating system | Windows, macOS, or Linux |
| Node.js | Version 24 or newer, required by the developer preview |
| Browser | Chrome, Edge, Brave, or Firefox for the hosted workspace. Safari is unverified; the local address works in every browser |
| Disk space | Space for YveChat Engine, Ollama, and every model you download |
| Network | Required to download software and models, and to load the hosted interface |
YveChat Engine listens on a loopback address on your machine. YveChat Engine and the runtimes it talks to run on the same computer.
Chromium browsers and Firefox are verified to reach the Engine from `app.yvechat.xyz`. Safari has historically blocked a website from reaching loopback and remains unverified; a Mac user who cannot connect can open `http://127.0.0.1:3210` directly, which works in every browser. A future release removes the Node.js requirement by bundling its own runtime.
***
### Memory and model size
Model files load into memory before they produce output. The size of the model determines how much memory the machine needs.
| Model scale | Typical use | Practical guidance |
| ------------------------------------ | --------------------------------------------------------- | ------------------------------------------------------------------------- |
| Small, roughly 1B to 3B parameters | Quick replies, simple drafting, older or lighter machines | Runs on modest hardware and produces the lowest quality of the three |
| Medium, roughly 7B to 14B parameters | General conversation and everyday work | The common starting point on a modern machine with adequate memory |
| Large, 30B parameters and above | Demanding reasoning and long documents | Requires substantial memory and benefits significantly from a capable GPU |
Sizes above describe general categories. Actual memory consumption varies by model, quantization, and configured context length, so treat any figure as an estimate and test on your own hardware.
{% hint style="info" %}
Universal compatibility is not claimed. Some combinations of model, machine, and operating system will perform poorly or fail to load. Starting with a small model establishes whether the setup works before you commit to a large download.
{% endhint %}
***
### Quantization
Quantization reduces the precision of a model's weights so the file occupies less memory. A more heavily quantized model loads on smaller machines and produces lower quality output. Ollama documents quantization as a tradeoff between memory and precision, with the effect becoming more noticeable at longer context lengths.
Source: [Ollama FAQ](https://docs.ollama.com/faq)
***
### Graphics acceleration
A supported GPU increases speed substantially. Ollama documents NVIDIA support at compute capability 5.0 and above with specified minimum driver versions, along with AMD ROCm on Linux, Metal on Apple hardware, and Vulkan on Windows and Linux.
Source: [Ollama GPU documentation](https://docs.ollama.com/gpu)
Machines without a supported GPU fall back to the processor. Generation still works and runs considerably slower.
***
### Context length and memory
Ollama defaults to a 4096-token context window, and required memory scales with the number of parallel requests multiplied by the context length. Raising the context window to handle longer documents increases memory consumption proportionally.
Source: [Ollama FAQ](https://docs.ollama.com/faq)
***
### Image generation
Local image generation runs through ComfyUI and carries its own requirements, which are generally heavier than language models of comparable popularity. Image models benefit from a dedicated GPU with substantial video memory.
***
### What running locally consumes
Local inference uses your electricity, your memory, your storage, and your processing time. A long generation on a laptop drains the battery and warms the machine. Large models occupy tens of gigabytes of disk each.
Those costs are the price of custody, and they are yours to carry.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:
```
GET https://yvechat.gitbook.io/wp/getting-started/system-requirements.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.