Raon-1 nano (LiteRT-LM)

Raon-1 nano (LiteRT-LM)

Packs Raon-1 nano as a 4-bit LiteRT-LM bundle with a 4,096-token context, for running a Korean and English chat model on devices. Replies come without a thinking step.

Made by KRAFTON on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: This repository holds a LiteRT-LM bundle of Raon-1 nano, a compact KRAFTON AI language model for Korean and English chat. The bundle stores weights in 4 bits using GPTQ, runs activations in fp16 and has a context of 4,096 tokens. It replies directly, with no thinking phase.

Who it suits: It suits developers who build on-device apps with LiteRT-LM and want a small Korean and English model.

Title: Who might use it
Industry | Who uses it | How they use it | Example
Mobile apps | App developers | Ship a small chat model inside a device app | An in-app assistant that answers in Korean or English
Edge devices | Embedded engineers | Try a 4-bit model where memory is tight | A trial of on-device text replies
AI research | Researchers | Compare the 4-bit bundle with the full model | Checking what the short context window changes

People who need long context or a thinking step can skip it, since the full model supports 262,144 tokens.

Where to use it:

• LiteRT-LM: the library this bundle is made for
• Windows, macOS or Linux: download the files with Python
• Other runtimes: the card lists Transformers, vLLM, llama.cpp and MLX for Raon-1 nano, and this bundle is the LiteRT-LM one

Title: Jobs it does
Use case | Who | What they get back
On-device chat | App developers | Short replies in Korean or English from a 4-bit bundle
Quick answers | Product teams | Direct answers with no thinking step
Instruction following | Developers | Text written to match the instruction given

How to set it up: You need these first:

• Python on Windows, macOS or Linux
• A LiteRT-LM setup that can load the bundle

1. Install the Hugging Face hub package.
2. Download the bundle files.
3. Load the bundle in LiteRT-LM, following the model card.

pip install -U huggingface_hub
hf download KRAFTON/Raon-1-nano-litert-lm

How to use it: Download the bundle, load it in LiteRT-LM and send prompts in Korean or English. This command fetches the files.

hf download KRAFTON/Raon-1-nano-litert-lm

The card lists sampling settings by task for Raon-1 nano.

Title: Recommended sampling settings
Task | Temperature | Presence penalty
General chat, knowledge questions, instruction following | 0.1 | 1.5
Agentic work or code | 1.0 | Not set

The output token limit is worth raising for longer replies.

Example: You load the bundle, set temperature to 0.1 and presence penalty to 1.5, and ask a general question in Korean. A direct text reply comes back with no thinking text in front of it.

For newcomers and experts: If you are new to this, think of it as a small chat model packed for phones and other devices. Download it and try one short question once your LiteRT-LM setup is ready.

If you work in the field, the base model is a hybrid with 24 layers, 18 of them linear attention and 6 full attention. This bundle quantizes the weights to 4 bits with GPTQ and keeps activations in fp16. Its 4,096-token window is far shorter than the 262,144 tokens of the full model.

Download the bundle
Load it in LiteRT-LM
Send a Korean or English prompt
Read the direct reply

Good to know: The context window is the main trade-off against the full model.

Title: Context length in tokens
This bundle: 4096 tokens
Full Raon-1 nano: 262144 tokens

• License: apache-2.0
• Cost to run: none named
• Data it reaches: only the prompts you give it, with no outside service named
• Not for: prompts longer than its 4,096-token context

On Hugging Face, 13 November 2026

• Task: text generation
• Library: litert-lm
• Downloads in the last 30 days: 26
• Likes: 12
• License: apache-2.0
• Based on: KRAFTON/Raon-1-nano
• Tags: litert-lm, raon, text-generation

Links

• Model: https://huggingface.co/KRAFTON/Raon-1-nano-litert-lm
• Files: https://huggingface.co/KRAFTON/Raon-1-nano-litert-lm/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download KRAFTON/Raon-1-nano-litert-lm

Then follow the usage in its model card. Nothing in this kit runs by itself.