What are the hardware requirements for Fireball 12B?

Fireball 12B requires approximately 24.5 GB of VRAM and supports a context window of 1000K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Fireball 12B and how large is it?

Fireball 12B is developed by EpistemeAI, a model with 12b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate Fireball 12B?

Fireball 12B is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Fireball 12B by EpistemeAI — VRAM 24.5GB, 1000K context

Name: Fireball 12B
Rating: 4 (4 reviews)
Author: EpistemeAI

Fireball 12B is an open-source language model by EpistemeAI. Features: 12b LLM, VRAM: 24.5GB, Context: 1000K, License: apache-2.0, LLM Explorer Score: 0.18.

Base model:epistemeai/fireball... Base model:finetune:epistemeai... Dataset:candenizkocak/code-alp... Dataset:reciperesearch/dolphin... Dataset:yahma/alpaca-cleaned En Endpoints compatible Mistral Model-index Region:us Safetensors Sharded Tensorflow Trl Unsloth

Model Card on HF 🤗: https://huggingface.co/EpistemeAI/Fireball-12B

Fireball 12B Parameters and Internals

Model Type

text-generation-inference, transformers, unsloth, mistral, trl

Additional Notes

Model is a pretrained base model and does not have any moderation mechanisms.

Training Details

Data Sources:

candenizkocak/code-alpaca-297k, yahma/alpaca-cleaned, reciperesearch/dolphin-sft-v0.1-preference

Methodology:

Supervised fine-tuning

Context Length:

128000

Model Architecture:

Transformer model with 40 layers, 5,120 dimensionality, 128 head dim, 14,436 hidden dim, SwiGLU activation, 32 heads, 8 kv-heads

LLM Name	Fireball 12B
Repository 🤗	https://huggingface.co/EpistemeAI/Fireball-12B
Base Model(s)	EpistemeAI/Fireball-Mistral-Nemo-Base-2407-sft-v2.2a EpistemeAI/Fireball-Mistral-Nemo-Base-2407-sft-v2.2a
Model Size	12b
Required VRAM	24.5 GB
Updated	2026-08-06
Maintainer	EpistemeAI
Model Type	mistral
Model Files	4.9 GB: 1-of-5 4.9 GB: 2-of-5 4.9 GB: 3-of-5 4.9 GB: 4-of-5 4.9 GB: 5-of-5
Supported Languages	en
Model Architecture	MistralForCausalLM
License	apache-2.0
Context Length	1024000
Model Max Length	1024000
Transformers Version	4.44.0
Tokenizer Class	PreTrainedTokenizerFast
Padding Token	<pad>
Vocabulary Size	131072
Torch Data Type	bfloat16

Rank the Fireball 12B 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 Fireball 12B

Best Alternatives	Context / RAM	Downloads	Likes
...r Nemo 12B Instruct R 21 09 24	1000K / 24.5 GB	396277	141
MN 12B Mag Mell R1	1000K / 24.5 GB	69558	238
Riverfish Rocinante 12B SFT	1000K / 49.4 GB	166	1
Riverfish Rocinante 12B	1000K / 49.4 GB	111	1
MN 12B LucidFaun RP RU	1000K / 24.5 GB	1005	6
Qliphoth 12B V1.2	1000K / 24.5 GB	107	7
Captain Eris BMO Violent 12B	1000K / 24.5 GB	305	2
Dans SakuraKaze V1.0.0 12B	1000K / 24.5 GB	7	21
Shining Gold Sakura 12B	1000K / 24.5 GB	52	2
...n Eris BMO Violent GRPO V0.420	1000K / 24.5 GB	92	3

Note: green Score (e.g. "73.2") means that the model is better than EpistemeAI/Fireball-12B.