Meta has introduced Llama 3, the latest in its series of large language models, boasting significant advancements in AI capabilities. The new models, based on the Llama 2 architecture, come in two sizes – 8 billion (8B) and 70 billion (70B) parameters, each offering a base model and an instruction-tuned version for specific tasks like powering chatbots for user conversations.
Powering Meta AI, a new assistant on Facebook, Instagram, WhatsApp, and Messenger, Llama 3 models support a context length of 8,000 tokens, allowing for more extended interactions and complex input handling. Integrated into the Hugging Face ecosystem, the models are readily available to developers through tools like transformers and inference endpoints, as well as model-as-a-service providers and cloud platforms.
In addition to the Llama 3 models, Meta has released Llama Guard 2, a safety model fine-tuned on the 8B version to improve production cases’ safety and reliability. With impressive performance across various benchmarks, the 70B model outperforms high-profile models like OpenAI’s GPT-3.5 and Google’s Gemini in tasks such as coding, creative writing, and summarization.
Trained on a dataset comprising 15 trillion tokens, about seven times larger than that used for Llama 2, the extensive training has significantly contributed to the models’ improved performance and capabilities. Meta plans to develop more capable versions of Llama 3, exceeding 400 billion parameters, to support multiple languages and modalities, enhancing the model’s versatility and applicability.
Emphasizing its commitment to the open-source community, Meta offers Llama 3 for free, fostering innovation and allowing for widespread testing and improvement by developers globally. Optimized for hardware from Intel, AMD, and NVIDIA, Intel has published a detailed guide on the model’s performance on its Gaudi AI accelerators and Xeon CPUs. Llama 3 is expected to gain widespread adoption with Meta’s reach and partnerships with major industry players.
Overall, Llama 3 from Meta represents a significant advancement in open models, with its advanced capabilities, support for various tasks, and commitment to the open-source community. With partnerships with major industry players and a growing adoption rate, Llama 3 is expected to continue evolving and gaining popularity in the AI industry.
