What are the hardware requirements for Rocket 3B?

Rocket 3B requires approximately 5.6 GB of VRAM and supports a context window of 4K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Rocket 3B and how large is it?

Rocket 3B is developed by pansophic, a model with 3b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Rocket 3B perform on standard benchmarks?

Rocket 3B has the following published scores: MMLU 47.1. Compare against reference models on this page or on the LLM Explorer leaderboards.

Rocket 3B by pansophic — VRAM 5.6GB, 4K context

Name: Rocket 3B
Author: pansophic

Rocket 3B is an open-source language model by pansophic. Features: 3b LLM, VRAM: 5.6GB, Context: 4K, License: cc-by-sa-4.0, LLM Explorer Score: 0.12, Arc: 50.6, HellaSwag: 76.7, MMLU: 47.1, GSM8K: 36.5.

Arxiv:2101.00027 Arxiv:2305.06161 Arxiv:2305.18290 Base model:finetune:stabilitya... Base model:stabilityai/stablel... En Endpoints compatible Pytorch Region:us Safetensors Stablelm

Model Card on HF 🤗: https://huggingface.co/pansophic/rocket-3B

Rocket 3B Parameters and Internals

Model Type

GPT-like, text generation

Use Cases

Areas:

chat, conversation

Additional Notes

The model name is inspired by Rocket Raccoon from 'Guardians of the Galaxy'.

Supported Languages

English (Primarily)

Training Details

Data Sources:

openbmb/UltraFeedback, BAAI/JudgeLM-100K

Methodology:

Direct Preference Optimization (DPO)

Input Output

Input Format:

ChatML format

Accepted Modalities:

text

Output Format:

text response

LLM Name	Rocket 3B
Repository 🤗	https://huggingface.co/pansophic/rocket-3B
Base Model(s)	Stablelm 3B 4e1t stabilityai/stablelm-3b-4e1t
Model Size	3b
Required VRAM	5.6 GB
Updated	2026-07-28
Maintainer	pansophic
Model Type	stablelm
Model Files	5.6 GB 5.6 GB
Supported Languages	en
Model Architecture	StableLmForCausalLM
License	cc-by-sa-4.0
Context Length	4096
Model Max Length	4096
Transformers Version	4.38.0
Tokenizer Class	GPTNeoXTokenizer
Vocabulary Size	50304
Torch Data Type	bfloat16

Rank the Rocket 3B 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

Quantized Models of the Rocket 3B

Model	Likes	Downloads	VRAM
Rocket 3B GGUF	40	3324	1 GB
Rocket 3B GPTQ	0	8	1 GB

Best Alternatives to Rocket 3B

Best Alternatives	Context / RAM	Downloads	Likes
...t 3B Mix Spider Bird 200 Steps	16K / 5.6 GB	9	0
... Instruct 3B Spider 3500 Steps	16K / 11.2 GB	5	0
Stablelm Zephyr 3B	4K / 5.6 GB	25834	261
Stablelm 3B 4e1t	4K / 5.6 GB	77091	315
Stablelm 3B Abliterated	4K / 5.6 GB	6	1
ReMask 3B	4K / 11.2 GB	71	16
Stablelm 4e1t 2B V0.1	4K / 4 GB	6	0
Machroom 3B Model Stock	4K / 5.6 GB	106	0
Zephyr Sumbot All Songs	4K / 5.6 GB	13	1
Canvers Slm Ov V1	4K / 2.8 GB	9	0

Note: green Score (e.g. "73.2") means that the model is better than pansophic/rocket-3B.