LLM EXPLORER 59,420 MODELS INDEXED

Llama 3.1 405B Instruct by meta-llama

By meta-llama · 43229 downloads
HF Repo ↗ Copy URL

Llama 3.1 405B Instruct is an open-source language model by meta-llama. Features: 405b LLM, VRAM: 183.1GB, Context: 128K, License: llama3.1, Instruction-Based, LLM Explorer Score: 0.21.

  Arxiv:2204.05149 Base model:finetune:meta-llama... Base model:meta-llama/llama-3....   Conversational   De   En   Endpoints compatible   Es   Eval-results   Facebook   Fr   Hi   Instruct   It   Llama   Llama-3   Meta   Pt   Pytorch   Region:us   Safetensors   Sharded   Tensorflow   Th
Model Card on HF 🤗: https://huggingface.co/meta-llama/Llama-3.1-405B-Instruct 

Llama 3.1 405B Instruct Parameters and Internals

Model Type 
text-generation, multilingual
Use Cases 
Areas:
commercial, research
Applications:
assistant-like chat, natural language generation tasks, synthetic data generation
Primary Use Cases:
multilingual dialogue
Limitations:
not for use beyond 8 supported languages without fine-tuning and compliance with terms
Considerations:
ensure safety in additional languages
Additional Notes 
Used for commercial and research purposes. Regular updates to improve model safety with community feedback.
Supported Languages 
English (full), German (full), French (full), Italian (full), Portuguese (full), Hindi (full), Spanish (full), Thai (full)
Training Details 
Data Sources:
publicly available online data
Data Volume:
15 trillion tokens
Methodology:
supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF)
Context Length:
128000
Training Time:
39.3M GPU hours
Hardware Used:
custom built GPU cluster, H100-80GB
Model Architecture:
optimized transformer architecture
Safety Evaluation 
Methodologies:
red teaming, risk assessments, evaluation datasets
Findings:
some safety risks identified and mitigated
Risk Categories:
misinformation, cyber threats, child safety
Ethical Considerations:
engagement strategies with subject-matter experts for real-world harms
Responsible Ai Considerations 
Fairness:
efforts to mitigate bias through multi-faceted data collection approach
Transparency:
part of an open community for AI safety progress
Accountability:
use of output reporting mechanism
Mitigation Strategies:
adopting MLCommons taxonomy, employing numerous safety guardrails
Input Output 
Input Format:
multilingual text
Accepted Modalities:
text
Output Format:
multilingual text and code
Performance Tips:
Follow the Responsible Use Guide
Release Notes 
Version:
3.1
Date:
2024-07-23
Notes:
Enhancements for multilingual dialogue, improved benchmarks results.
LLM NameLlama 3.1 405B Instruct
Repository 🤗https://huggingface.co/meta-llama/Llama-3.1-405B-Instruct 
Base Model(s)  meta-llama/Meta-Llama-3.1-405B   meta-llama/Meta-Llama-3.1-405B
Model Size405b
Required VRAM183.1 GB
Updated2026-08-07
Maintainermeta-llama
Model Typellama
Instruction-BasedYes
Model Files  4.8 GB: 1-of-191   4.0 GB: 2-of-191   4.6 GB: 3-of-191   4.6 GB: 4-of-191   3.5 GB: 5-of-191   4.6 GB: 6-of-191   4.6 GB: 7-of-191   3.5 GB: 8-of-191   4.6 GB: 9-of-191   4.6 GB: 10-of-191   3.5 GB: 11-of-191   4.6 GB: 12-of-191   4.6 GB: 13-of-191   3.5 GB: 14-of-191   4.6 GB: 15-of-191   4.6 GB: 16-of-191   3.5 GB: 17-of-191   4.6 GB: 18-of-191   4.6 GB: 19-of-191   3.5 GB: 20-of-191   4.6 GB: 21-of-191   4.6 GB: 22-of-191   3.5 GB: 23-of-191   4.6 GB: 24-of-191   4.6 GB: 25-of-191   3.5 GB: 26-of-191   4.6 GB: 27-of-191   4.6 GB: 28-of-191   3.5 GB: 29-of-191   4.6 GB: 30-of-191   4.6 GB: 31-of-191   3.5 GB: 32-of-191   4.6 GB: 33-of-191   4.6 GB: 34-of-191   3.5 GB: 35-of-191   4.6 GB: 36-of-191   4.6 GB: 37-of-191   3.5 GB: 38-of-191   4.6 GB: 39-of-191   4.6 GB: 40-of-191   3.5 GB: 41-of-191   4.6 GB: 42-of-191   4.6 GB: 43-of-191
Supported Languagesen de fr it pt hi es th
Model ArchitectureLlamaForCausalLM
Licensellama3.1
Context Length131072
Model Max Length131072
Transformers Version4.43.3
Vocabulary Size128256
Torch Data Typebfloat16

Best Alternatives to Llama 3.1 405B Instruct

Best Alternatives
Context / RAM
Downloads
Likes
...ta Llama 3.1 405B Instruct FP8128K / 197.6 GB55370165
Llama 3.1 405B Instruct FP8128K / 209.2 GB476016
Meta Llama 3.1 405B Instruct128K / 186 GB55654473
Llama 3.1 405B Instruct FP8128K / 193.4 GB15734193
Shisa V2 Llama3.1 405B128K / 191.2 GB2219
BigLlama 3.1 681B Instruct128K / 189.6 GB1712
INTELLECT 1 Instruct8K / 20.5 GB328125
...ama 3.1 405B Instruct Bnb 4bit128K / 214.3 GB1419
Note: green Score (e.g. "73.2") means that the model is better than meta-llama/Llama-3.1-405B-Instruct.