# Run AI Models Locally: Llama.cpp's New Web Interface Makes It Dead Simple

**URL:** <https://onehack.st/t/run-ai-models-locally-llama-cpps-new-web-interface-makes-it-dead-simple/312696>\
**Category:** News & Articles\
**Tags:** tools, programming, privacy, ai\
**Created:** [November 5, 2025, 11:21am UTC](https://onehack.st/t/run-ai-models-locally-llama-cpps-new-web-interface-makes-it-dead-simple/312696 "2025-11-05T11:21:41Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![Shayla](https://onehack.st/user_avatar/onehack.st/shayla/32/174870_2.png) [@Shayla](https://onehack.st/u/Shayla)\
**Post date:** [November 5, 2025, 11:21am UTC](https://onehack.st/t/run-ai-models-locally-llama-cpps-new-web-interface-makes-it-dead-simple/312696/1 "2025-11-05T11:21:41Z")

</div>

# 🦙 Llama.cpp’s Web UI Just Graduated — No More Preview Mode

> **🗺 One-Line Flow:**  
> After months in preview, **llama.cpp** now ships a **stable, built-in web UI** — clean, fast, and finally ready for humans.

## For 🫏 1Hackers

- Don’t worry, this is the update that finally lets you play “ChatGPT” on your own PC — no cloud, no login, just pure nerd magic. ⚡

![nicksplat rugrats GIF](https://media1.giphy.com/media/v1.Y2lkPTNiMWQxM2I2Zndqd2hrdmFicWtpemtyM2t5aHVrZzJ3MzVxYzBodjd6djB5MTlnMiZlcD12MV9naWZzX3NlYXJjaCZjdD1n/l1JoiYABTH0iRyrS0/giphy.gif)

* * *

### ⚙ What’s New

The new interface is **SvelteKit-based** , rewritten by **Alek (allozaur)** — one of llama.cpp’s core devs.  
It’s a full migration from the old React preview that started back in **September 2025 (PR #14839)**.  
Now, in **November 2025** , it’s officially stable — no more nightly builds or half-broken previews.

## 👇 Watch it in action

> <https://x.com/TheAhmadOsman/status/1985751771638350061/video/1>

* * *

### 🧩 Killer Features

Here’s why it’s not just another chat box:

- 💬 **RAG mode:** Upload text, PDFs, or images — it chats with your docs directly.
- 🧠 **Smart paste:** Long text? It auto-saves as a file.
- 👁 **Vision models:** Drop an image and it _actually understands it_.
- 📂 **File manager:** Upload, preview, convert (SVG, WEBP, text — all supported).
- 🧵 **Chat branching:** Conversations split into trees — no lost context.
- ⚡ **Shortcuts:**  
`Ctrl+Shift+N` → New chat  
`Ctrl+K` → Search  
`Ctrl+B` → Toggle sidebar
- 📑 **PDF dual view:** Text extraction + side-by-side pages.

Basically, it’s **ChatGPT-style polish** , but local and open-source.

![bunny GIF](https://media1.giphy.com/media/v1.Y2lkPTNiMWQxM2I2cXE5bHR0a3RwMmU5ZDMzOHdkN3RybW9kNzR0c3ZheWZhZWtvNndxZCZlcD12MV9naWZzX3NlYXJjaCZjdD1n/bDC9Px6yjuiE8/giphy.gif)

* * *

### 🧰 How to Run

1. Update to the latest llama.cpp build.

2. Launch the server:

3. Open your browser → `127.0.0.1:8080`

That’s it. No dependencies. No cloud. Just vibes.  
Docs: [llama.cpp on GitHub](https://github.com/ggml-org/llama.cpp)

 ![image](https://onehack.st/uploads/default/original/3X/b/d/bd0c51915e3560b685730ef4bd3afab064430b08.jpeg)

* * *

### ⚖ Ollama vs Llama.cpp

- **Llama.cpp Web UI:**  
Lightweight, local, plug-and-play.  
Great for quick chats and RAG workflows.

- **Ollama:**  
Handles model downloads, loading/unloading, and Modelfiles better.  
Perfect for managing multiple models or complex setups.

Both are great — pick your poison.

* * *

### 🧨 Quick Notes

Some users say it’s missing **multi-agent handling** — true.  
For that, check **ExLlama** or **vLLM** (they handle parallel workloads better).

* * *

### (ಠ‿ಠ) Girllll… another AI update? Fine. But can it _pay rent_ this time?

_Because if not, I’m installing it on my toaster and calling it a startup._

![Not Bad Season 3 GIF by The Real Bros of Simi Valley](https://media2.giphy.com/media/v1.Y2lkPTNiMWQxM2I2djg3eXNzbHB4ZG1ndjcydzJjYWw5Y2w0ZW9hNG5rczk0dWVsd3JqYSZlcD12MV9naWZzX3NlYXJjaCZjdD1n/dZS6VBPvhoJhkdi1RG/giphy.gif)

* * *

1. **Local ChatGPT Clone Seller**

* * *

1. **Digital Product Farm**

* * *

1. **Privacy Freaks Market**

* * *

1. **Custom PDF Chat Bots**

* * *

1. **Resume & Cover Letter Factory**

* * *

1. **Café / Library Setup**

* * *

1. **The Pretend Hacker Route**

* * *

💀 **In short:**  
llama.cpp’s Web UI isn’t just a nerd toy — it’s an ATM in disguise.  
All it takes is a bit of imagination and one command:  
`./llama-server`

* * *

### 💀 In Short

The web UI isn’t “new,” it’s just **finally done right**.  
Alek rebuilt it from scratch, made it fast, stable, and smart.  
Now llama.cpp isn’t just for coders — it’s for everyone.

---

<div class="post-metadata">

**Author:** ![gulraizrana](https://onehack.st/user_avatar/onehack.st/gulraizrana/32/148856_2.png) [@gulraizrana](https://onehack.st/u/gulraizrana)\
**Post date:** [November 12, 2025, 10:01am UTC](https://onehack.st/t/run-ai-models-locally-llama-cpps-new-web-interface-makes-it-dead-simple/312696/2 "2025-11-12T10:01:57Z")

</div>

hi, how much disk space, CPU and RAM needed for this?

---

<div class="post-metadata">

**Author:** ![Parizaad](https://onehack.st/user_avatar/onehack.st/parizaad/32/174871_2.png) [@Parizaad](https://onehack.st/u/Parizaad)\
**Post date:** [November 12, 2025, 10:10am UTC](https://onehack.st/t/run-ai-models-locally-llama-cpps-new-web-interface-makes-it-dead-simple/312696/3 "2025-11-12T10:10:13Z")

</div>

Small model? Chill — 8 GB RAM and 10 GB space got you.  
Big brain 13B? Aim for 16–32 GB RAM.  
Wanna flex with 70B? Bring 64 GB + and a beefy GPU — or watch your CPU cry.