LLM EXPLORER 58,975 MODELS INDEXED

Normistral 7B Scratch by norallm

By norallm · 192 downloads
HF Repo ↗ Copy URL

Normistral 7B Scratch is an open-source language model by norallm. Features: 7b LLM, VRAM: 14.5GB, Context: 2K, License: apache-2.0, Quantized, LLM Explorer Score: 0.11.

  Arxiv:2005.14165   Arxiv:2204.02311   Arxiv:2302.01398   Dataset:nbailab/ncc   Dataset:uonlp/culturax Dataset:vikp/starcoder filtere...   Endpoints compatible   Generative   Gguf   Gpt   Mistral   Nb   Nn   No   Pytorch   Quantized   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/norallm/normistral-7b-scratch 

Normistral 7B Scratch Parameters and Internals

Model Type 
text generation
Use Cases 
Areas:
research
Considerations:
The model is primarily intended for research purposes and may generate harmful completions from inappropriate prompts.
Supported Languages 
Norwegian Bokmål (native), Norwegian Nynorsk (native)
Training Details 
Data Sources:
NCC corpus, HPLT corpus, CulturaX, Starcoder
Data Volume:
260 billion subword tokens
Methodology:
pretrained from scratch
Context Length:
2000
Hardware Used:
LUMI supercomputer with AMD MI250X GPUs
Model Architecture:
auto-regressive transformer architecture based on Mistral/Llama models
Input Output 
Input Format:
Textual input.
Accepted Modalities:
text
Output Format:
Generated text.
LLM NameNormistral 7B Scratch
Repository 🤗https://huggingface.co/norallm/normistral-7b-scratch 
Model Size7b
Required VRAM14.5 GB
Updated2026-07-31
Maintainernorallm
Model Typemistral
Model Files  4.8 GB: 1-of-3   4.8 GB: 2-of-3   4.9 GB: 3-of-3   4.4 GB   5.1 GB   6.0 GB   7.7 GB   14.5 GB
Supported Languagesno nb nn
GGUF QuantizationYes
Quantization Typegguf
Model ArchitectureMistralForCausalLM
Licenseapache-2.0
Context Length2048
Model Max Length2048
Transformers Version4.34.0.dev0
Tokenizer ClassPreTrainedTokenizerFast
Vocabulary Size32768
Torch Data Typebfloat16

Best Alternatives to Normistral 7B Scratch

Best Alternatives
Context / RAM
Downloads
Likes
MegaBeam Mistral 7B 300K Gguf282K / 5 GB353
MegaBeam Mistral 7B 300K Gguf282K / 5 GB343
Backrooms Mistral 7B 10e32K / 14.5 GB1280
Mistral 7B Instruct V0.2.gguf32K / 14.5 GB173
Mistral 7B Instruct V0.2 GGUF32K / 4.4 GB262
...andle Mistral 7B Instruct V0.232K / 14.4 GB1030
...andle Dolphin 2.2.1 Mistral 7B32K / 14.4 GB800
Boptruth Agatha 7B32K / 14.4 GB1290
Mahou 1.2a Mistral 7B32K / 14.4 GB337
Tamil Mistral 7B Instruct V0.132K / 14.8 GB35014
Note: green Score (e.g. "73.2") means that the model is better than norallm/normistral-7b-scratch.