What are the hardware requirements for BigLlama 3.1 681B Instruct?

BigLlama 3.1 681B Instruct requires approximately 189.6 GB of VRAM and supports a context window of 128K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed BigLlama 3.1 681B Instruct and how large is it?

BigLlama 3.1 681B Instruct is developed by mlabonne, a model with 405b 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 BigLlama 3.1 681B Instruct?

BigLlama 3.1 681B Instruct is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

BigLlama 3.1 681B Instruct by mlabonne — VRAM 189.6GB, 128K context

Name: BigLlama 3.1 681B Instruct
Author: mlabonne

BigLlama 3.1 681B Instruct is an open-source language model by mlabonne. Features: 405b LLM, VRAM: 189.6GB, Context: 128K, Instruction-Based, Merged, LLM Explorer Score: 0.14.

Merged Model Base model:finetune:meta-llama... Base model:meta-llama/llama-3.... Conversational Endpoints compatible Instruct Llama Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/mlabonne/BigLlama-3.1-681B-Instruct

BigLlama 3.1 681B Instruct Parameters and Internals

Model Type

text generation

Use Cases

Applications:

creative writing

Additional Notes

This is an experimental self-merge using Meta-Llama-3.1-405B-Instruct and was created with the use of mergekit. It is suggested to use the Llama 3 chat template for optimal performance.

LLM Name	BigLlama 3.1 681B Instruct
Repository 🤗	https://huggingface.co/mlabonne/BigLlama-3.1-681B-Instruct
Base Model(s)	meta-llama/Meta-Llama-3.1-405B-Instruct meta-llama/Meta-Llama-3.1-405B-Instruct
Merged Model	Yes
Model Size	405b
Required VRAM	189.6 GB
Updated	2026-08-02
Maintainer	mlabonne
Model Type	llama
Instruction-Based	Yes
Model Files	4.2 GB: 1-of-322 4.2 GB: 2-of-322 3.5 GB: 3-of-322 4.7 GB: 4-of-322 4.7 GB: 5-of-322 3.5 GB: 6-of-322 4.7 GB: 7-of-322 3.5 GB: 8-of-322 3.5 GB: 9-of-322 4.2 GB: 10-of-322 3.5 GB: 11-of-322 3.5 GB: 12-of-322 4.7 GB: 13-of-322 4.7 GB: 14-of-322 3.5 GB: 15-of-322 4.7 GB: 16-of-322 4.7 GB: 17-of-322 3.5 GB: 18-of-322 4.7 GB: 19-of-322 4.7 GB: 20-of-322 3.5 GB: 21-of-322 4.7 GB: 22-of-322 4.7 GB: 23-of-322 4.7 GB: 24-of-322 4.7 GB: 25-of-322 3.5 GB: 26-of-322 4.7 GB: 27-of-322 4.7 GB: 28-of-322 3.5 GB: 29-of-322 4.7 GB: 61-of-322 3.5 GB: 62-of-322 4.7 GB: 63-of-322 4.7 GB: 64-of-322 3.5 GB: 65-of-322 4.7 GB: 66-of-322 4.7 GB: 67-of-322 3.5 GB: 68-of-322 4.7 GB: 69-of-322 4.7 GB: 70-of-322 3.5 GB: 71-of-322 3.5 GB: 72-of-322 4.7 GB: 73-of-322 4.7 GB: 74-of-322 3.5 GB: 75-of-322 4.7 GB: 76-of-322
Model Architecture	LlamaForCausalLM
Context Length	131072
Model Max Length	131072
Transformers Version	4.43.3
Vocabulary Size	128256
Torch Data Type	bfloat16

Rank the BigLlama 3.1 681B Instruct 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 BigLlama 3.1 681B Instruct

Best Alternatives	Context / RAM	Downloads	Likes
...ta Llama 3.1 405B Instruct FP8	128K / 197.6 GB	55370	165
Llama 3.1 405B Instruct FP8	128K / 209.2 GB	4760	16
Llama 3.1 405B Instruct	128K / 183.1 GB	43229	595
Meta Llama 3.1 405B Instruct	128K / 186 GB	55654	473
Llama 3.1 405B Instruct FP8	128K / 193.4 GB	15734	193
Shisa V2 Llama3.1 405B	128K / 191.2 GB	22	19
INTELLECT 1 Instruct	8K / 20.5 GB	328	125
...ama 3.1 405B Instruct Bnb 4bit	128K / 214.3 GB	141	9

Note: green Score (e.g. "73.2") means that the model is better than mlabonne/BigLlama-3.1-681B-Instruct.