⚡️ Llama 3.3 70B.
The model is available in a version with 70 billion parameters and is optimized for conversational use cases in multiple languages. Llama 3.3 outperforms many available open and closed source models on industry standard benchmarks.
Llama 3.3 is based on an optimized Transformer architecture and uses an autoregressive approach. Model tuning includes SFT with RLHF to align with human preferences for utility and safety.
The model has been trained on a new publicly available online dataset of over 15 trillion tokens, with a data freshness constraint until December 2023.
Llama 3.3 supports English, German, French, Italian, Portuguese, Hindi, Spanish and Thai

⚡️ Llama 3.3 70B.

The model is available in a version with 70 billion parameters and is optimized for conversational use cases in multiple languages. Llama 3.3 outperforms many available open and closed source models on industry standard benchmarks.

Llama 3.3 is based on an optimized Transformer architecture and uses an autoregressive approach. Model tuning includes SFT with RLHF to align with human preferences for utility and safety.

The model has been trained on a new publicly available online dataset of over 15 trillion tokens, with a data freshness constraint until December 2023.

Llama 3.3 supports English, German, French, Italian, Portuguese, Hindi, Spanish and Thai

Meta released Llama 3.3 70 b :Best Open Source llm

Table of contents

No headings in the article.