Llama 3.3 70B
Overview
Llama 3.3 70B is a large language model developed by Meta, released in December 2024. It features 70 billion parameters and utilizes an optimized transformer architecture, employing techniques such as supervised fine-tuning and reinforcement learning with human feedback to enhance instruction-following capabilities. The model supports a context window of 128,000 tokens, allowing for extensive input processing. It is available under the Llama 3 Community License, enabling open access for research and commercial applications. Designed for text-based tasks, Llama 3.3 70B demonstrates improved performance in reasoning, coding, mathematics, and instruction adherence compared to its predecessor, Llama 3.1 70B. It has been benchmarked on various tasks, achieving an MMLU score of 86.0, a HumanEval pass rate of 88.4%, and a Math score of 77.0%. These results indicate its proficiency in handling complex language understanding and generation tasks. The model is accessible through multiple providers, including Microsoft Azure AI Foundry, Cerebras, Google Gemini, Google Vertex AI, Snowflake, and Vercel AI Gateway, with pricing varying by provider.