What are the hardware requirements for Open Llama 3B V2?

Open Llama 3B V2 requires approximately 6.8 GB of VRAM and supports a context window of 2K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Open Llama 3B V2 and how large is it?

Open Llama 3B V2 is developed by openlm-research, 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 Open Llama 3B V2 perform on standard benchmarks?

Open Llama 3B V2 has the following published scores: MMLU 27.3. Compare against reference models on this page or on the LLM Explorer leaderboards.

Open Llama 3B V2 by openlm-research — VRAM 6.8GB, 2K context

Name: Open Llama 3B V2
Author: openlm-research

Open Llama 3B V2 is an open-source language model by openlm-research. Features: 3b LLM, VRAM: 6.8GB, Context: 2K, License: apache-2.0, LLM Explorer Score: 0.17, Arc: 46.6, HellaSwag: 68.2, MMLU: 27.3, GSM8K: 3.1.

Arxiv:2302.13971 Dataset:bigcode/starcoderdata Dataset:tiiuae/falcon-refinedw... Dataset:togethercomputer/redpa... Deploy:azure Endpoints compatible Llama Pytorch Region:us

Model Card on HF 🤗: https://huggingface.co/openlm-research/open_llama_3b_v2

Open Llama 3b V2 Benchmarks

ARC: 46.6 vs 96.7 (so35)^-51.8%

HellaSwag: 68.2 vs 95.3 (gpt4)^-28.4%

MMLU: 27.3 vs 88.3 (so35)^-69.1%

GSM8K: 3.1 vs 96.4 (so35)^-96.8%

OpenCompas: 34.6 vs 73.3 (gpt4)^-52.8%

Exam: 31 vs 77.2 (gpt4)^-59.8%

Language: 33.2 vs 62 (gpt4)^-46.5%

Knowledge: 42.1 vs 73.5 (gpt4)^-42.7%

Understand: 33.2 vs 70 (gpt4)^-52.6%

Reasoning: 34.4 vs 74.4 (gpt4)^-53.8%

LLME Score: 0.16827

^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)

Open Llama 3B V2 (openlm-research/open_llama_3b_v2)

🌟 Advertise your project 🚀

Open Llama 3B V2 Parameters and Internals

Model Type

large language model

Additional Notes

Please note that it is advised to avoid using the Hugging Face fast tokenizer for now, as it sometimes gives incorrect tokenizations. This can be avoided by using 'use_fast=False'.

Training Details

Data Sources:

tiiuae/falcon-refinedweb, bigcode/starcoderdata, togethercomputer/RedPajama-Data-1T

Data Volume:

1 trillion tokens

Methodology:

Pre-trained with open datasets rather than the original LLaMA dataset, using the EasyLM framework.

Hardware Used:

cloud TPU-v4s

LLM Name	Open Llama 3b V2
Repository 🤗	https://huggingface.co/openlm-research/open_llama_3b_v2
Model Size	3b
Required VRAM	6.8 GB
Updated	2026-07-05
Maintainer	openlm-research
Model Type	llama
Model Files	6.8 GB
Model Architecture	LlamaForCausalLM
License	apache-2.0
Context Length	2048
Model Max Length	2048
Transformers Version	4.31.0.dev0
Tokenizer Class	LlamaTokenizer
Beginning of Sentence Token	<s>
End of Sentence Token	</s>
Unk Token	<unk>
Vocabulary Size	32000
Torch Data Type	float16

Quantized Models of the Open Llama 3B V2

Model	Likes	Downloads	VRAM
...3b V2 Python Instruct 0.1 4bit	0	7	2 GB

Best Alternatives to Open Llama 3B V2

Best Alternatives	Context / RAM	Downloads	Likes
Nanbeige4.1 3B	256K / 7.9 GB	11138	1142
Nanbeige4.1 3B Heretic	256K / 7.9 GB	149	50
Nanbeige4.1 3B Bf16	256K / 7.9 GB	47	3
Nanbeige4.1 3B Heretic Mxfp4	256K / 7.9 GB	33	2
Nanbeige4.1 3B Heretic	256K / 7.9 GB	5	1
Nanbeige4.1 3B Heretic	256K / 7.9 GB	5	0
ISA 03 Mini 3B Hybrid Preview	256K / 6.5 GB	851	4
Llama 3.2 3B Instruct	128K / 6.5 GB	2018457	2279
Llama 3.2 3B	128K / 6.5 GB	706937	845
... V7 SFT 10voicebot Loopcleaned	128K / 6.5 GB	221	0

Note: green Score (e.g. "73.2") means that the model is better than openlm-research/open_llama_3b_v2.

Rank the Open Llama 3B V2 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? 54964 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

Open Llama 3B V2 by openlm-research