First Look: Meta Llama 2 Open Source Large Language Model

Unlock this article and more with a membership!

Become a Member
First Look: Meta Llama 2 Open Source Large Language Model
This is some text inside of a div block.

The New Chapter in Open Source: The LLM Story Continues

Llama 2 Graphic by Jim Intriglia/Midjourney
Meta Llama 2 (Courtesy JiE/Midjourney

As an Open-Source advocate, the announcement of the release of Meta's Llama 2 large language model caught my attention. The 1991 decision by Linus Torvalds to Open Source his Linux operating system kernel marked a significant milestone in the history of business computing. Making the source code of the Linux Kernel available to the worldwide community of software developers laid the groundwork for developing the platform that powers the majority of business computing back-end platforms.

Meta's decision to release Llama 2 as an Open-Source product will significantly benefit business enterprises of all sizes, worldwide. This decision lays the groundwork for developing a freely available LLM platform that will level the playing field for entrepreneurs, developers and business enterprises worldwide.

Introduction to Llama 2

Llama 2 is the second version of the open-source language model from Meta. It is based on a transformer architecture and has now also been released for commercial use. This presents many new possibilities, as a variety of applications can be built on the Llama architecture.

Llama stands for Large Language Model Meta AI. It is an auto-regressive large language model that uses an optimized transformer architecture. It is the second Foundational Model from Meta AI, which was released in 2023. The first version of Llama, released in late February 2023, was already open source. Llama 2 is now not only completely Open-Source, but it can also be used commercially as well.

Key Features and Capabilities

Llama 2 comes in different sizes with 7, 13, 34, or 70 billion parameters. It was trained on 2 trillion tokens with over a million human annotations. Context length has doubled from version 1 to 2, from about 2000 to about 4000 tokens. Besides the size variants, there’s also a fine-tuned variant of the model for chat applications called Llama 2-Chat.

Performance and Benchmarks

Compared to other open-source models, like Falcon-40B or Mosaic ML's MPT, Meta's new model performs excellently. This moves it to the top stop in the Hugging Face Open LLM Leaderboard. However, compared to proprietary models, like OpenAI’s GPT models, there’s still room for improvement. In some benchmarks, Llama 2 outperforms GPT-3.5, the model based on ChatGPT. For example, in the Hella Swag benchmark, LLaMA-70B-chat performs better than GPT-3.5.

Training and Development

Llama 2 was initially trained in the pre-training phase using publicly available online sources, with a much larger dataset than Llama 1. After pre-training, the first version of Llama 2 chat was created through supervised fine-tuning. Human experts helped with the training at this stage. To further improve the model’s performance and create more natural responses, Reinforcement Learning from Human Feedback (RLHF) was implemented in the next step. This involves iteratively refining the model with reinforcement learning and human feedback.

Partnership with Microsoft and Future Prospects

At the same time as the Llama 2 release, Microsoft announced itis expanding its partnership with Meta. Starting today, Llama 2 is available in the Azure AI model catalog, enabling developers using Microsoft Azure to build with it and leverage their cloud-native tools for content filtering and safety features. It is also optimized to run locally on Windows, giving developers a seamless workflow as they bring generative AI experiences to customers across different platforms. Llama 2 is available through Amazon Web Services (AWS), Hugging Face, and other providers.

Responsible Use and Safety Measures

Meta has put exploratory research, open source, and collaboration with academic and industry partners at the heart of their AI efforts. They have also created new initiatives to harness the insight and creativity of individuals, researchers, and developers around the world to get feedback on how the models are performing and how they might be improved. They have also put a policy in place that prohibits certain use cases to help ensure that these models are being used fairly and responsibly.

Trying Llama 2 Chat via Free Web Platforms

You can try the Llama 2 chatbot on the following platforms:

 1. Web LLM. This project brings large-language model and LLM-based chatbot to web browsers. Llama 2 7B/13B are now available in Web LLM.

2. Chat with Open Large Language Models.  This platform allows you to chat with various models including Llama 2.

The availability and performance of these platforms can vary. Some web platforms may require you to create an account or provide certain permissions. Always ensure you're comfortable with the terms of use before interacting with online platforms.


Llama 2: New Open-Source Language Model from Meta Released. (Alexander Goschin, TechMio Blog, 7/21/2023).

Meta and Microsoft Introduce the Next Generation of Llama. (Meta, 7/18/2023)