In a technological landscape where innovation is king, the Llama models have emerged as a beacon of progress. With downloads approaching a staggering 350 million, these open-source models have not only captured the imagination of developers worldwide but have also set new benchmarks in the AI industry. The past year alone has seen a tenfold increase in downloads, with over 20 million downloads recorded in the last month, solidifying Llama's position as the leading open-source model family.
The release of Llama 3.1 has been a game-changer. Expanding the context length to 128K, supporting eight languages, and introducing the frontier-level Llama 3.1 405B model, this update has doubled the token volume usage across major cloud service providers from May to July 2024. The monthly token volume usage of Llama has grown tenfold from January to July 2024 for some of the largest cloud service providers, a testament to its growing utility and popularity.
Open-source has always been the cornerstone of Llama's success. By making these models openly available, a vibrant and diverse AI ecosystem has flourished, providing developers with unprecedented choice and capability. From start-ups pushing the boundaries of innovation to enterprises of all sizes integrating Llama into their operations, the impact has been broad and rapid. The industry is not just building with Llama; it's innovating at an unprecedented pace.
An open letter from a prominent tech leader further underscored the benefits of open-source AI, cementing the vision and commitment to an open approach. Open-source is in the DNA of the company behind Llama, and the models embody this commitment to responsible sharing. Open-source promotes a competitive ecosystem, benefiting consumers, companies, and society at large.
In just 18 months since its initial launch, Llama has evolved from a single state-of-the-art foundation model to a robust system for developers. Llama 3.1 offers a complete reference system, enabling developers to create custom agents with ease. New security and safety tools have also been introduced, ensuring responsible development.
The Llama ecosystem is expanding at a breakneck pace. With downloads on platforms like Hugging Face approaching 350 million, an over tenfold increase from last year, the momentum is undeniable. In August alone, Llama models were downloaded more than 20 million times on Hugging Face. This success is mirrored across various industry partners, including AWS, Microsoft Azure, Databricks, Dell, Google Cloud, Groq, NVIDIA, IBM watsonx, Scale AI, Snowflake, and others.
Hosted Llama usage by token volume across major cloud service providers more than doubled from May to July 2024. The highest number of unique users of Llama 3.1 in August was for the 405B variant, indicating that the largest foundation model is gaining significant traction. The number of partners in the Llama early access program has grown fivefold with Llama 3.1, and the demand from potential partners continues to surge.
Industry experts have lauded the impact of Llama. One expert from AWS highlighted the excitement surrounding Llama 3.1, noting that customers are eager to use the latest state-of-the-art models for building AI applications in the cloud. Another expert from Databricks emphasized that thousands of customers have adopted Llama 3.1, making it the fastest adopted and best-selling open-source model in their history. The sentiment is echoed by an expert from Groq, who noted that the demand for Llama is so high that even a tenfold increase in deployed capacity would be consumed in under 36 hours.
NVIDIA's CEO also praised Llama, stating that it has profoundly impacted the advancement of state-of-the-art AI. The rapid pace of adoption is a testament to the model's capabilities and the industry's readiness to embrace open-source AI.
The developer community's growing preference for Llama is evident. According to a survey from Artificial Analysis, Llama was the second most considered model and the industry leader in open-source. With over 60,000 derivative models on Hugging Face, developers are fine-tuning Llama for various use cases. Large enterprises like AT&T, DoorDash, Goldman Sachs, Niantic, Nomura, Shopify, Spotify, and Zoom are just a few success stories, with Infosys and KPMG using Llama internally.
Accenture is leveraging Llama 3.1 to build a custom LLM for ESG reporting, expecting to improve productivity by 70% and quality by 20-30%. The multilingual capabilities of Llama 3.1 are helping global organizations make chatbots more culturally conscious and relevant. AT&T has fine-tuned Llama models to improve customer care, driving a nearly 33% improvement in search-related responses while reducing costs and speeding up response times.
DoorDash uses Llama to streamline tasks for its software engineers, leveraging its internal knowledge base to answer complex questions and improve its codebase. Goldman Sachs uses Llama models for information extraction from documents, while Niantic has integrated Llama into its AR game Peridot, transforming virtual pets into responsive, smart creatures.
Nomura uses Llama on AWS for text summarization, code generation, log analysis, and document processing. Shopify experiments with LLaVA, built on Llama, for product metadata and enrichment, performing 40M-60M inferences per day. Zoom uses Llama to power its AI Companion, helping users with meeting summaries, smart recordings, and next steps.
Llama is leading the way in openness, modifiability, and cost efficiency. The commitment to building in the open ensures that the benefits of AI extend to everyone. As the Llama ecosystem grows, so do the capabilities and accessibility of Meta AI. The smart assistant is now available across Instagram, WhatsApp, Messenger, Facebook, Meta Quest, and the Ray-Ban Meta collection, bringing the vision of a contextual AI assistant closer to reality.
The growth of the Llama community is a testament to the model's impact. Developers and companies are finding immense value in Llama, and the future looks bright. Stay tuned for more updates as the Llama journey continues.
For those new to Llama, the latest models are available for download, offering a gateway to building the next generation of AI applications. The open-source revolution is here, and Llama is leading the charge.