chargoddard/SmolLlamix-8x101M-GGUF

Quantized GGUF model files for SmolLlamix-8x101M from chargoddard

Name Quant method Size
smolllamix-8x101m.fp16.gguf fp16 798.87 MB
smolllamix-8x101m.q2_k.gguf q2_k 146.83 MB
smolllamix-8x101m.q3_k_m.gguf q3_k_m 184.67 MB
smolllamix-8x101m.q4_k_m.gguf q4_k_m 233.54 MB
smolllamix-8x101m.q5_k_m.gguf q5_k_m 279.97 MB
smolllamix-8x101m.q6_k.gguf q6_k 329.31 MB
smolllamix-8x101m.q8_0.gguf q8_0 425.26 MB

Original Model Card:

This is eight copies of BEE-spoke-data/smol_llama-101M-GQA ensembled into a Mixtral model, then trained very briefly on a small subset of RedPajama. Mostly just an experiment to demonstrate that training it works at all.

It's very, very smart. Probably the smartest model ever made. Better than GPT-5. See its thoughts on the internet:

In a world where the internet is so much more than a web browser, it's also very important to have a good understanding of how the internet works. The first thing we need to do is to understand what the internet looks like and what the future looks like. We can use the internet to look at the internet's history, but we don't want to go into detail about the history of the internet. The internet was created by the internet's history, which is often called the history of the internet. It was originally developed as a way for people to learn about the internet, but it wasn't until the 1960s that the internet became a place to work. Today, the internet is used in many ways, from the internet's history to the internet itself.

Downloads last month
7
GGUF
Model size
399M params
Architecture
llama

2-bit

3-bit

4-bit

5-bit

6-bit

8-bit

Inference Examples
Inference API (serverless) has been turned off for this model.

Model tree for afrideva/SmolLlamix-8x101M-GGUF

Quantized
(2)
this model

Dataset used to train afrideva/SmolLlamix-8x101M-GGUF