Meta has launched Llama 3, its most advanced large language model (LLM) to date. The new model is open-source and includes improved reasoning abilities and higher performance metrics. Llama 3 can be accessed through popular platforms like AWS or Google Cloud permitting developers and researchers to leverage its capabilities, per the report.
Also read: Samsung announces major AI investment in smartphones
Meta unveiled two versions of Llama 3 – an 8-billion and 70-billion parameter version, which are both significantly better than previous models. The company highlights its improved transformer architecture, training regime, and instruction tuning as key factors that contribute to these advancements.
🔵 META UNVEILS BIGGEST LLAMA 3 AI MODEL, TOUTING LANGUAGE AND MATH GAINS
— PiQ (@PiQSuite) July 23, 2024
Meta Platforms released the biggest version of its mostly free Llama 3 artificial intelligence models on Tuesday, boasting multilingual skills and general performance metrics that nip at the heels of paid… pic.twitter.com/5vVUBx5d1T
Meta Llama 3 model brings new capabilities and multilingual support
Llama 3 is built to excel in different applications and can work well in brainstorming sessions, classification tasks, creative writing prompts, and answering open-ended queries. Meta ran many human evaluation tests of Llama 3 against other models such as Claude Sonnet, Mistral Medium, and OpenAI’s GPT-3.5. Although specific comparisons to OpenAI’s GPT-4 were not provided, Meta said that Llama 3 performs better than previous versions in real-world situations.
The main characteristics of Llama 3 are its additional coding skills and improved reasoning skills. In future updates, Meta plans to introduce multimodal features and languages. Meta aims to encourage creativity and cooperation among AI developers by making the model open-source. This will help in advancing AI as a whole by allowing developers to suggest improvements and take part in fine-tuning the model driving the next wave of AI advancements.
Also read: Greece begins AI training for public servants
Meta’s open-source approach with Llama 3 follows their last launch which was the Llama 2 in 2023. Even though some people have criticized them for restricting developers who work for rival companies and also because of safety filters that may be present in this particular model, Meta remains committed to transparency and community engagement. The firm emphasizes the importance of ethical AI development and governance, aiming to balance innovation with responsible practices.
The introduction of Llama 3 was praised by Iris.ai Chief Technology Officer Victor Botev, who mentioned that open-sourcing models for AI development are very important. He also recognized what Meta has done in terms of ethics and governance concerning artificial intelligence but called for sustained commitment from all stakeholders.
