What are the hardware requirements for Raccoon Small?

Raccoon Small requires approximately 38.4 GB of VRAM and supports a context window of 32K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Raccoon Small and how large is it?

Raccoon Small is developed by rizla, a model with 19.2b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Raccoon Small perform on standard benchmarks?

Raccoon Small has the following published scores: MMLU 64.55. Compare against reference models on this page or on the LLM Explorer leaderboards.

Raccoon Small by rizla — VRAM 38.4GB, 32K context

Name: Raccoon Small
Author: rizla

Raccoon Small is an open-source language model by rizla. Features: 19.2b LLM, VRAM: 38.4GB, Context: 32K, License: cc-by-nc-4.0, MoE, Instruction-Based, LLM Explorer Score: 0.11, Arc: 74.4, HellaSwag: 88.7, MMLU: 64.6, GSM8K: 56.9.

Base model:finetune:mistralai/... Base model:mistralai/mixtral-8... Dpo Endpoints compatible Instruct Mixtral Moe Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/rizla/raccoon-small

Raccoon Small Parameters and Internals

Model Type

language model

Additional Notes

Made specifically as part of an experimental merging process using the mergekitty tool.

Training Details

Methodology:

Merged using mergekitty tool; Based on GPT-4 architecture.

Model Architecture:

Combination of two 2expmixtrals into a singular model via merging.

LLM Name	Raccoon Small
Repository 🤗	https://huggingface.co/rizla/raccoon-small
Base Model(s)	mistralai/Mixtral-8x7B-Instruct-v0.1 mistralai/Mixtral-8x7B-Instruct-v0.1
Model Size	19.2b
Required VRAM	38.4 GB
Updated	2026-07-21
Maintainer	rizla
Model Type	mixtral
Instruction-Based	Yes
Model Files	10.0 GB: 1-of-4 10.0 GB: 2-of-4 9.9 GB: 3-of-4 8.5 GB: 4-of-4
Model Architecture	MixtralForCausalLM
License	cc-by-nc-4.0
Context Length	32768
Model Max Length	32768
Transformers Version	4.37.2
Tokenizer Class	LlamaTokenizer
Padding Token	<s>
Vocabulary Size	32000
Torch Data Type	bfloat16

Rank the Raccoon Small capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to Raccoon Small

Best Alternatives	Context / RAM	Downloads	Likes
...oundary Solar Chat 2x10.7B MoE	4K / 38 GB	123	1
SOLAR Math 2x10.7B V0.2	4K / 38.4 GB	212	4

Note: green Score (e.g. "73.2") means that the model is better than rizla/raccoon-small.