What are the hardware requirements for Llama 3B Finnish?

Llama 3B Finnish requires approximately 7.3 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 Llama 3B Finnish and how large is it?

Llama 3B Finnish is developed by Finnish-NLP, 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 3B Finnish?

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

Llama 3B Finnish by Finnish-NLP — VRAM 7.3GB, 2K context

Name: Llama 3B Finnish
Author: Finnish-NLP

Llama 3B Finnish is an open-source language model by Finnish-NLP. Features: 3b LLM, VRAM: 7.3GB, Context: 2K, License: apache-2.0, LLM Explorer Score: 0.07.

Arxiv:2302.06675 Arxiv:2302.13971 Arxiv:2305.16264 Dataset:finnish-nlp/mc4 3.1.0 ... Dataset:finnish-nlp/oscar 2301... Dataset:finnish-nlp/reddit fi ... Dataset:finnish-nlp/wikipedia ... Dataset:intfloat/multilingual ... Fi Finnish Llama Pytorch Region:us Safetensors

Model Card on HF 🤗: https://huggingface.co/Finnish-NLP/llama-3b-finnish

Llama 3B Finnish Parameters and Internals

Model Type

text generation

Use Cases

Areas:

Research, Commercial applications

Primary Use Cases:

Text generation

Limitations:

Can have biased predictions, Better to fine-tune for specific tasks

Additional Notes

Pretrained on a very large corpus of Finnish data without human labeling. Recommended to fine-tune for better results.

Supported Languages

Finnish (high)

Training Details

Data Sources:

Finnish-NLP/mc4_3.1.0_fi_cleaned, Finnish-NLP/oscar_2301_fi_cleaned, Finnish-NLP/Reddit_fi_2006_2022, Finnish-NLP/wikipedia_20230501_fi_cleaned, Yle Finnish News Archive 2011-2018, Yle Finnish News Archive 2019-2020, Finnish News Agency Archive (STT), The Suomi24 Sentences Corpus, Project Lönnrot, Finnish parliament speeches, multilingual_cc_news, fi-news-corpus

Data Volume:

19 billion words, 28 billion tokens

Methodology:

Self-supervised causal language modeling

Context Length:

2048

Training Time:

131B tokens

Hardware Used:

TPUv4-32 VM

Model Architecture:

Pretrained Llama model on Finnish language using a causal language modeling objective.

Responsible Ai Considerations

Fairness:

Model may have biases due to content from the internet.

Mitigation Strategies:

Toxic content filtered with a toxicity classifier but cannot eliminate all.

Input Output

Input Format:

Sequences of continuous text

Accepted Modalities:

Text

Output Format:

Text

Performance Tips:

Best for generating texts from a prompt.

LLM Name	Llama 3B Finnish
Repository 🤗	https://huggingface.co/Finnish-NLP/llama-3b-finnish
Model Size	3b
Required VRAM	7.3 GB
Updated	2026-07-28
Maintainer	Finnish-NLP
Model Type	llama
Model Files	7.3 GB 7.3 GB
Supported Languages	fi
Model Architecture	LlamaForCausalLM
License	apache-2.0
Context Length	2048
Model Max Length	2048
Transformers Version	4.34.0.dev0
Tokenizer Class	LlamaTokenizer
Beginning of Sentence Token	<s>
End of Sentence Token	</s>
Unk Token	<unk>
Vocabulary Size	64256
Torch Data Type	float16

Rank the Llama 3B Finnish 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 Llama 3B Finnish

Best Alternatives	Context / RAM	Downloads	Likes
Nanbeige4.1 3B	256K / 7.9 GB	20875	1161
Nanbeige4.1 3B Heretic	256K / 7.9 GB	149	50
Nanbeige4.1 3B Heretic Mxfp4	256K / 7.9 GB	34	2
Nanbeige4.1 3B Bf16	256K / 7.9 GB	47	3
...anbeige4.1 3B Heretic Bfloat16	256K / 7.9 GB	33	1
Nanbeige4.1 3B Heretic	256K / 7.9 GB	14	1
Nanbeige4.1 3B Heretic	256K / 7.9 GB	15	0
ISA 03 Mini 3B Hybrid Preview	256K / 6.5 GB	4	4
ToolMind Web 3B	192K / 7.9 GB	156	28
Llama 3.2 3B Instruct	128K / 6.5 GB	1829572	2368

Note: green Score (e.g. "73.2") means that the model is better than Finnish-NLP/llama-3b-finnish.