Llama

Llama is a foundation-model family from Meta designed for AI developers to build custom applications without vendor lock-in.

Highlights

• Model variety: 8B, 70B, and 405B parameter sizes let you optimize for speed, accuracy, or resource constraints
• Context capacity: up to 128K tokens in recent releases supports document summarization and multi-turn conversations
• Instruction-tuned versions for following complex directions, reasoning, and multi-step problem-solving
• Strong code generation, reasoning, and multilingual performance across benchmarks
• Completely free, open-source, and commercially usable for both research and production systems

Whether you're building conversational agents, content understanding systems, or specialized AI tools, Llama models let you deploy on your infrastructure or leverage affordable cloud providers. The open nature makes it accessible for startups and enterprises who need control over their AI systems and want to avoid being locked into proprietary platforms.