Llama 3.3 70B — Meta

Llama 3.3 70B is Meta's large instruct model: an excellent balance of quality and speed for writing, everyday reasoning, summarization and code. With quality close to far heavier models, it stays fast and multilingual.

Type : Text · Parameters : 70 billion · Context : 128K tokens

Strengths

• Writing and summarization
• Multilingual (including French)
• Great quality/speed ratio

Best for

• Everyday general-purpose assistant
• Document summarization
• Drafting emails and articles

Llama 3.3 70B is served via NVIDIA NIM, HuggingFace, included in your subscription with dozens of other models — one price, every model.