Llama 3.2 (1B & 3B) #115

Model description

Llama 3.2 (1B & 3B) are lightweight text-only models optimized for edge devices, supporting a 128k context length.

Open source examples / papers / references

• Paper: The Llama 3 Herd of Models (covers Llama 3 architecture)
• Official Release: Llama 3.2 Announcement
• Website: Llama 3.2 Documentation
• Model code: Hugging Face Transformers (LlamaModel)
• Model weights:
  • 1B / 1B-Instruct
  • 3B / 3B-Instruct

Why this model?

Llama 3.2 offers strong performance in the 1B/3B parameter class. Its low memory footprint enables LLM research on consumer-grade hardware.