Meta Releases Llama 4: The Most Capable Open-Source Model Yet

Meta has released Llama 4, the latest generation of its open-source large language model family. The flagship 405-billion parameter variant represents a major leap in capability, with benchmark scores placing it in direct competition with top closed-source models from OpenAI and Anthropic.

Mark Zuckerberg announced the release on his social channels, calling it "the most powerful open model ever released" and emphasizing Meta's commitment to keeping frontier AI accessible to researchers, developers, and businesses worldwide.

Llama 4 introduces a new architecture with improved long-context handling (supporting up to 128K tokens), enhanced multilingual support across 30+ languages, and significantly better performance on coding, mathematics, and structured reasoning tasks.

The model weights are available on Hugging Face and Meta's website under a commercial license that allows use for most applications. Smaller variants at 8B, 70B, and 405B parameters cater to different compute budgets.

The open-source AI community has responded enthusiastically, with fine-tuned versions already appearing within hours of the release. Analysts view the launch as a strategic move to keep the AI ecosystem open and to grow Meta's developer ecosystem ahead of future product integrations.