Llama 3
Llama 3 is Meta's open-weight large language model family released in 8B and 70B parameter sizes, trained on over 15 trillion tokens using 24,000 GPUs with an 8,192 token context window. The model has been widely adopted across the open source AI ecosystem for diverse applications, from vision-language model training to agentic systems, and recent research has identified architectural trade-offs in the model that impact long context performance.
Updated 3 August 2026
Specifications
No specifications recorded yet.
Latest developments
OlmPool: How small architectural choices compound to undermine long context extension
Allen Institute (AI2) · 3 months ago ·
10
The Frontier is Open
Together AI · 1 year ago ·
30
nanoVLM: The simplest repository to train your VLM in pure PyTorch
Hugging Face Blog · 1 year ago ·
34
License to Call: Introducing Transformers Agents 2.0
Hugging Face Blog · 2 years ago ·
38
Data Machina #251
Data Machina · 2 years ago ·
34
Data Machina #250
Data Machina · 2 years ago ·
26
Welcome Llama 3 - Meta's new open LLM
Hugging Face Blog · 2 years ago ·
15
2026
2025
2024
Relationships
Products & technology
- Meta develops this model · 2 sources
- Derived from Llama 2 · 1 source
- Integrated with Hugging Face · 1 source
- Integrated with Google Cloud · 1 source
- Integrated with Amazon SageMaker · 1 source