LLaMA 3

LLaMA 3 is a free, open-weight large language model from Meta, first released on July 23, 2024, as the successor to Llama 2. It launched in two primary sizes (8B and 70B parameters), with a 405B variant following later, and was trained on over 15 trillion tokens—roughly seven times the data used for Llama 2. The model introduces grouped query attention for improved inference speed and delivers a measured score of 63.0 on the MMLU-Pro benchmark, along with 81.7% pass@1 on HumanEval. It is offered at zero cost for both input and output tokens ($0/1M input, $0/1M output), making it a no-fee option for text generation and reasoning tasks. LLaMA 3 sits within a broader Llama family that includes successors such as Llama 3.1, Llama 3.3, and Llama 4, underscoring that it is an earlier, stable release rather than the cutting edge. Its significance lies in combining verified benchmark performance with genuinely free access, providing a practical baseline for evaluating cost-free, open-source language models in research and enterprise settings.

Two-hop subgraph: this entity, every entity it directly relates to, and every entity those neighbors relate to. Drag a node, scroll to zoom, click to inspect — or click any neighbor and re-center the atlas there.