What are the hardware requirements for Llammas?

Llammas requires approximately 13.5 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 Llammas and how large is it?

Llammas is developed by tartuNLP, a model with 6.7b 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 Llammas?

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

Llammas by tartuNLP — VRAM 13.5GB, 4K context

Name: Llammas
Author: tartuNLP

Llammas is an open-source language model by tartuNLP. Features: 6.7b LLM, VRAM: 13.5GB, Context: 4K, License: llama2, LLM Explorer Score: 0.1.

Arxiv:2404.04042 Base model:finetune:tartunlp/l... Base model:tartunlp/llammas-ba... Conversational Deploy:azure En Endpoints compatible Et Eval-results Llama Pytorch Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/tartuNLP/Llammas

Llammas Parameters and Internals

Model Type

instruction-tuned, text-generation, conversational

Additional Notes

It is the first open-source instruction-following LLM for Estonian.

Supported Languages

Estonian (high proficiency), English (high proficiency)

Training Details

Data Sources:

CulturaX, Alpaca-cleaned, Alpaca-est, OASST1 top-1 English conversations, CoT, FLAN-V2, WMT18 English-Estonian translation development data, general MTee validation English-Estonian held-out data

Data Volume:

5B tokens

Methodology:

Continued pre-training and instruction-tuning

Model Architecture:

LLama-based architecture

Input Output

Input Format:

Conversational format with roles for user and assistant

Accepted Modalities:

text

Output Format:

Generated text based on conversation history

Performance Tips:

Recommended to use with transformers version 4.36.2 for best output

LLM Name	Llammas
Repository 🤗	https://huggingface.co/tartuNLP/Llammas
Base Model(s)	Llammas Base tartuNLP/Llammas-base
Model Size	6.7b
Required VRAM	13.5 GB
Updated	2026-08-08
Maintainer	tartuNLP
Model Type	llama
Model Files	10.0 GB: 1-of-2 3.5 GB: 2-of-2 10.0 GB: 1-of-2 3.5 GB: 2-of-2
Supported Languages	et en
Model Architecture	LlamaForCausalLM
License	llama2
Context Length	4096
Model Max Length	4096
Transformers Version	4.34.0.dev0
Tokenizer Class	LlamaTokenizer
Beginning of Sentence Token	<s>
End of Sentence Token	</s>
Unk Token	<unk>
Vocabulary Size	32000
Torch Data Type	bfloat16

Rank the Llammas 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 Llammas

Best Alternatives	Context / RAM	Downloads	Likes
Gptnoise37	128K / 13.5 GB	67	0
...Deepseek Coder 6.7B Energy Sft	16K / 13.5 GB	261	0
Deepseek Coder 6.7B Base	16K / 13.5 GB	54216	124
Speechless Coder Ds 6.7B	16K / 13.5 GB	459	7
...s Coder6.7b Reflct Adamw Iter2	16K / 13.5 GB	480	0
...ir8 Ds Coder6.7b Rmsprop Iter4	16K / 13.5 GB	183	0
...ir8 Ds Coder6.7b Rmsprop Iter2	16K / 13.5 GB	183	0
...s Coder6.7b Reflct Adamw Iter1	16K / 13.5 GB	475	0
...ir8 Ds Coder6.7b Rmsprop Iter3	16K / 13.5 GB	165	0
...s Coder6.7b Reflct Adamw Iter3	16K / 13.5 GB	430	0

Note: green Score (e.g. "73.2") means that the model is better than tartuNLP/Llammas.