What are the hardware requirements for Llama 3.2 3B?

Llama 3.2 3B requires approximately 6.5 GB of VRAM and supports a context window of 128K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Llama 3.2 3B and how large is it?

Llama 3.2 3B is developed by meta-llama, a model with 3b 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 Llama 3.2 3B?

Llama 3.2 3B is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Llama 3.2 3B by meta-llama — VRAM 6.5GB, 128K context

Name: Llama 3.2 3B
Author: meta-llama

Llama 3.2 3B is an open-source language model by meta-llama. Features: 3b LLM, VRAM: 6.5GB, Context: 128K, License: llama3.2, LLM Explorer Score: 0.32.

Model Card on HF 🤗: https://huggingface.co/meta-llama/Llama-3.2-3B

Llama 3.2 3B Benchmarks

IFEval: 13.37 vs 88 (so35)^-84.8%

MATH Lvl 5: 1.89

LLME Score: 0.32025

^nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

What is the LLM Explorer Rank (Score)

🌟 Advertise your project 🚀

Llama 3.2 3B Parameters and Internals

Model Type

text generation, text summarization, multilingual

Use Cases

Areas:

commercial, research

Applications:

knowledge retrieval, summarization, writing assistants, query and prompt rewriting

Primary Use Cases:

assistant chat, agentic applications

Limitations:

Use beyond supported languages without compliance with license, Violation of Acceptable Use Policy, Illegal activities

Considerations:

Encourage responsible use.

Additional Notes

Initial release for commercial and research purposes.

Supported Languages

en (English), de (German), fr (French), it (Italian), pt (Portuguese), hi (Hindi), es (Spanish), th (Thai)

Training Details

Data Volume:

9 trillion tokens

Methodology:

Supervised Fine-Tuning (SFT), Rejection Sampling (RS), Direct Preference Optimization (DPO).

Context Length:

128000

Training Time:

916k GPU hours

Hardware Used:

H100-80GB GPUs

Model Architecture:

Optimized transformer architecture with autoregressive language model.

Safety Evaluation

Methodologies:

Red teaming, Safety fine-tuning

Findings:

N/A

Risk Categories:

CBRNE

Responsible Ai Considerations

Mitigation Strategies:

System safeguards include Llama Guard, Prompt Guard, and Code Shield.

Input Output

Input Format:

Multilingual Text

Accepted Modalities:

text

Output Format:

Multilingual Text and code

LLM Name	Llama 3.2 3B
Repository 🤗	https://huggingface.co/meta-llama/Llama-3.2-3B
Model Size	3b
Required VRAM	6.5 GB
Updated	2026-04-23
Maintainer	meta-llama
Model Type	llama
Model Files	5.0 GB: 1-of-2 1.5 GB: 2-of-2
Supported Languages	en de fr it pt hi es th
Model Architecture	LlamaForCausalLM
License	llama3.2
Context Length	131072
Model Max Length	131072
Transformers Version	4.45.0.dev0
Tokenizer Class	PreTrainedTokenizerFast
Vocabulary Size	128256
Torch Data Type	bfloat16

Quantized Models of the Llama 3.2 3B

Model	Likes	Downloads	VRAM
Teuta	2	1153	6 GB
Llama 3.2 3B Bnb 4bit	22	12310	2 GB
Llm 3 2 Flux Prompt	15	1783	6 GB
Llama 3.2 3B COT	2	134	6 GB
Llama 3.2 3B GGUF	2	168	2 GB

Best Alternatives to Llama 3.2 3B

Best Alternatives	Context / RAM	Downloads	Likes
Nanbeige4.1 3B	256K / 7.9 GB	676572	1019
Nanbeige4.1 3B Heretic	256K / 7.9 GB	573	48
Nanbeige4.1 3B Bf16	256K / 7.9 GB	138	3
Nanbeige4.1 3B Heretic	256K / 7.9 GB	5	1
Nanbeige4.1 3B Heretic	256K / 7.9 GB	6	0
ISA 03 Mini 3B Hybrid Preview	256K / 6.5 GB	851	4
Llama 3.2 3B Instruct	128K / 6.5 GB	1674918	1724
Llama 3.2 3B Instruct	128K / 6.5 GB	1831	9
Hermes 3 Llama 3.2 3B	128K / 6.5 GB	11869	175
Llama 3.2 3B RP Toxic Fuse	128K / 6.4 GB	14	2

Note: green Score (e.g. "73.2") means that the model is better than meta-llama/Llama-3.2-3B.

Rank the Llama 3.2 3B Capabilities

🆘 Have you tried this model? Rate its performance. This feedback would greatly assist ML community in identifying the most suitable model for their needs. Your contribution really does make a difference! 🌟

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

What open-source LLMs or SLMs are you in search of? 53444 in total.

Email us: info@extractum.io. Our Privacy Policy | Terms and Conditions | Suggest an improvement.

Our Social Media →

Original data from HuggingFace, OpenCompass and various public git repos.

Check out Ag3ntum — our secure, self-hosted AI agent for server management.

Release v20260328a

Support LLM Explorer

Llama 3.2 3B by meta-llama