Model reference · open weights

Meta-Llama-3

LLMs meta-llama Text gen 2 builds Open, with conditions 1.7M dl/mo

Meta-Llama-3 is an open-weight language model from Meta. Meta-Llama-3-70B (BF16) weighs 141 GB; the smallest configuration that runs it is B200 180 GB.

Meta-Llama-3 is an 8.0B parameter instruction-tuned language model developed by Meta for text generation and dialogue use cases.
It operates in English and utilizes an optimized transformer architecture with Grouped-Query Attention.
The model is distributed under the llama3 license.

Summary of the meta-llama/Meta-Llama-3-8B-Instruct model card, 2026-10-01. The estimate below is for another build of the family.

Hugging Face ↗GitHub ↗

What it is

Released by	Meta
Released	2024-04-17
Parameters	8.0B
VRAM	141 GB for the weights

What it runs on

Memory and cards for Meta-Llama-3-70B (BF16)

141 GBweights, file size

762 MBruntime overhead, at least

How much memory each request adds isn't estimated yet for this architecture. The weights need at least the cards below, plus room for the context.

Card	Weights alone
RTX 3060 12 GB … H200 141 GB 11 smaller cards	does not fit
B200 180 GB	fits
8× H100 80 GB tensor parallel	does not fit
8× A100 80 GB tensor parallel	does not fit
8× H200 141 GB tensor parallel	does not fit

Builds

Sizes, precisions & builds

Build	Params	Precision	Weights	Smallest setup
Meta-Llama-3-8B-Instruct ↗	8.0B	BF16	16.1 GB	RTX 4090 24 GB
Meta-Llama-3-70B (above) ↗	70.6B	BF16	141 GB	B200 180 GB

How it works

How language models work

Details

Published under	meta-llama
Type	Language models
Task	Text gen
Runs with	transformers
Popularity	1.7M downloads / month
Weights	141 GB, Meta-Llama-3-70B (BF16, file size)

Languages