- Meta releases its largest open-source AI model, Llama 3.1 405B.
- Company aims to dominate generative AI market.
- Meta’s push for openness faces challenges in energy consumption and potential model biases.
Bigger, better Llama
Meta released its biggest open-source AI model yet, Llama 3.1 405B, boasting 405 billion parameters.
While not the absolute largest, it competes with leading proprietary models and offers capabilities like code analysis and document summarization in eight languages.
Llama 3.1 boasts a larger context window and the ability to integrate with third-party tools. Meta is also releasing smaller versions, 8B and 70B, aimed at general use like chatbots and code generation.
Meta hopes to encourage developers to use these models by offering reference systems, safety tools, and an upcoming API for customizing Llama.
Openness with limits
Meta updated Llama’s license to allow using its outputs for creating new AI models but retains control over deployment for high-traffic apps. This push for openness is part of Meta’s strategy to dominate the generative AI market.
Meta aims to become the go-to company for generative AI by giving away tools like Llama and fostering a developer community.
However, concerns remain about the model’s energy consumption and its potential to perpetuate biases from training data.
