L3-8B-Stheno-v3.3-32K-GGUF

L3-8B-Stheno-v3.3-32K-GGUF

Static GGUF files of Sao10K's Stheno 8B model in sizes from 3.3 GB to 16.2 GB. For people who run an 8B model on their own computer, under a non-commercial license.

Made by mradermacher 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 GGUF files that mradermacher produced from the Stheno 8B model by Sao10K. Each file trades size for quality, so you choose one that fits your memory. The model has 8B parameters and lists English. The card does not say what the base model was tuned to do.

Who it suits: It suits people who want an 8B model running on their own computer in a compact file, for non-commercial use.

Industry | Who uses it | How they use it | Example
Home use | Local AI users | Run an 8B model on their own computer | Load the Q4_K_M file of 5.0 GB in a GGUF app
Software | Developers | Test a quantized model on local hardware | Compare Q4_K_S and Q8_0 for speed and quality
Research | Quantization researchers | Study how quality changes between sizes | Run one prompt set against several quant types

If you need commercial use or the original full-precision weights, you can skip it.

Where to use it: You open the files in a program that reads GGUF, on Windows, macOS or Linux. The card points to TheBloke's READMEs for how to use GGUF files.

• A local app or runner that loads GGUF files
• Windows, macOS or Linux with the huggingface_hub package
• The separate i1 repository, if you prefer weighted imatrix quants

Use case | Who | What they get back
Local chat | Home users | A model that answers on their own machine
Size testing | Developers | The smallest file that still works for them
Imatrix quants | Enthusiasts | Weighted versions in the card's other repository

How to set it up: You need Python for the download step, a program that loads GGUF files, and enough disk space for the file you choose. Hugging Face estimates about 4 GB at 4-bit.

• Python on Windows, macOS or Linux
• The Hugging Face hub package
• A GGUF-capable program and 3.3 to 16.2 GB of free space for one quant

1. Install the Hugging Face hub package.
2. Download the files, or just the one quant you want.
3. Load that file in your GGUF program.

pip install -U huggingface_hub
hf download mradermacher/L3-8B-Stheno-v3.3-32K-GGUF

How to use it: Choose a quant from the list, load it in your GGUF program and chat. The repository holds many files, so check the sizes before you pull everything. The card says IQ types are often better than non-IQ types of similar size.

Tool | What it does
Q2_K, 3.3 GB | The smallest file
IQ3_S, 3.8 GB | The card says it beats the Q3_K types
Q3_K_M, 4.1 GB | A small file that the card says has lower quality
Q4_K_S, 4.8 GB | The card marks it as fast and recommended
Q4_K_M, 5.0 GB | Also fast and recommended
Q6_K, 6.7 GB | The card rates its quality as high
Q8_0, 8.6 GB | Fast, and the best quality among the quantized files
f16, 16.2 GB | Full 16-bit weights, which the card calls overkill

Example: You pick the Q4_K_M file at 5.0 GB, load it in a GGUF program and send a message. The model answers in text from your own machine.

For newcomers and experts: If you are new to this, a GGUF file is a model squeezed into a smaller size so it can run at home. A first small step is to pick the Q4_K_M file, which the card recommends.

If you know the field, these are static quants, listed by size and not strictly by quality. The card says weighted imatrix quants live in a separate i1 repository from the same author.

Sao10K's Stheno 8B model
Static GGUF quantization
Files in many sizes
You pick one that fits your memory
Your GGUF program runs it

Good to know: The files range from 3.3 GB to 16.2 GB. The card links to a graph and a write-up that compare lower-quality quant types.

Title: File size by quant type, in GB
Q2_K: 3.3 GB
IQ3_S: 3.8 GB
Q3_K_M: 4.1 GB
Q4_K_S: 4.8 GB
Q4_K_M: 5.0 GB
Q5_K_M: 5.8 GB
Q6_K: 6.7 GB
Q8_0: 8.6 GB

• License: cc-by-nc-4.0, which allows non-commercial use only
• Cost to run: no API key or paid service is named, and the files range from 3.3 GB to 16.2 GB
• Data it reaches: the download comes from Hugging Face, and the model works on what you give it on your own machine
• Not for: commercial use, since the license is non-commercial

On Hugging Face, 13 November 2026

• Library: transformers
• Parameters: 8B
• Weights: about 4 GB at 4-bit (estimate)
• Formats: GGUF
• Downloads in the last 30 days: 1,827
• Likes: 5
• License: cc-by-nc-4.0
• Languages: en
• Based on: Sao10K/L3-8B-Stheno-v3.3-32K
• Tags: transformers, gguf, conversational

Links

• Model: https://huggingface.co/mradermacher/L3-8B-Stheno-v3.3-32K-GGUF
• Files: https://huggingface.co/mradermacher/L3-8B-Stheno-v3.3-32K-GGUF/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download mradermacher/L3-8B-Stheno-v3.3-32K-GGUF

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