Meta says its Llama models were downloaded almost 350M times, are used by AT&T and others, and usage via cloud providers more than doubled from May to July 2024
we just published a bunch of updates on the adoption we're seeing. And yes, we have a lot more work to do on dev tools and resources which we're bringing online as quickly as we can. https://ai.meta.com/... Mark Zuckerberg / @zuck : Llama is growing even faster than I expected: almost 350M downloads (>20M in the last month!) and a 10x jump in monthly usage since the start of the year. Excited to share the next set of updates and Llama models soon. Thanks to everyone building with us 💪 Vishvanand Subramanian / @vishvanands : ["Is Enthusiasm For Meta's Llama Slowing Down?"] Wonder if @zuck's post is a response to this @theinformation headline. If so, great to hear that open source models are killing it 🚀 @aiatmeta : Open source AI is the way forward and Llama is leading the way. Today we're sharing a snapshot of how that's going with the adoption and use of Llama models across the industry. 🦙 Highlights • Llama is approaching 350M downloads on Hugging Face. … Connor Hayes / @conno_r : 🦙🦙🦙🦙🦙 It has been awesome to see an ecosystem start to form around llama as more developers and businesses are using it and supporting it with their tools. We shared more on this in a blog post today: https://ai.meta.com/... Ahmad Al-Dahle / @aaldahle : Llama is fast becoming the industry standard for open source AI w/ nearly 350M downloads to date and usage more than doubling in the last four months across our cloud partners. More on this momentum here: https://ai.meta.com/... We're committed to delivering open models that are leadership-class and provide developers the best possible foundation to build on. … X: @groqinc : Jonathan Ross, Founder & CEO, Groq: “Open-source wins. Meta is building the foundation of an open ecosystem that rivals the top closed models and at Groq we put them directly into the hands of the developers—a shared value that's been fundamental at Groq since our beginning. To date Groq has provided over 400,000 developers with 5 billion free tokens daily, using the Llama suite of models and our LPU Inference. It's a very exciting time and we're proud to be a part of that momentum. We can't add capacity fast enough for Llama. If we 10x'd the deployed capacity it would be consumed in under 36 hours.” @zoom : What makes the quality of Zoom AI Companion so good? It's our federated approach to AI ✨ AI Companion leverages a combination of proprietary models alongside both closed and open-source Large Language Models (LLMs), including the renowned @Meta Llama. This strategic blend Jonathan Ross / @jonathanross321 : 350,000,000 downloads of an LLM is nuts! How long did it take Linux to get to that number? @reach_vb : Congratulations @AIatMeta - it has been one of the most enjoyable collaborations in the past years! Thank you for your continued belief in democratising the Machine Learning! Looking forward to the next editions 🤗 P.S. The model downloads are already at 400M 🦙 Nik / @ns123abc : Zucc will keep doing this for as long as it takes to kneecap or bankrupt OpenAI. [image] @nvidiaaidev : 🎉We offer our congrats to @AIatMeta on reaching nearly 350M downloads of Llama. 🦙 From our CEO Jensen Huang: “Llama has profoundly impacted the advancement of state-of-the-art AI. The floodgates are now open for every enterprise and industry to build and deploy custom Llama Rat King / @mikeisaac : interesting blog from meta saying llamas, llamas everywhere https://ai.meta.com/... Yann LeCun / @ylecun : Meta's Llama🦙 has become the dominant platform in the AI ecosystem. An exploding number of companies large and small, startups, governments, and non-profits, are using to build new products and services. Universities, researchers, and engineers are improving Llama and Evan / @stockmktnewz : Mark Zuckerberg just said Facebook's $META Llama models are approaching 350 million downloads to date [image] LinkedIn: Ash Aggarwal : Nice update from the AI at Meta team! Customers want access to the latest state-of-the-art models for building AI applications in the cloud … Pierre Roux : 🚀 Llama is experiencing 10x growth and is a leading force in AI innovation. — 💥With the release of Llama 3.1 … Ahmad Al-Dahle : The growth of Llama in 2024 has been phenomenal. Llama models are fast becoming the industry standard for open source AI with nearly 350M downloads … Prashant Ratanchandani : Our Llama models continue to see incredible growth in the ecosystem, reaching 350 million downloads, a 10X increase compared to this time last year. …
Context & Ripple Effects
Meta’s adoption update follows the April release of Llama 3 models in 8B and 70B sizes, which paired model availability with a stated push to reduce false refusals. Meta was also extending Llama-powered Meta AI across its consumer apps, including Instagram, WhatsApp, Messenger, and Facebook.
The new figures matter because they add enterprise and cloud-provider usage signals to a distribution strategy that spans both Meta’s own products and third-party developers.
First-order effects
- Meta gains a stronger adoption proof point for Llama, while AT&T and the other named users have a publicly validated model option alongside proprietary AI services.
- More than doubling cloud-provider usage from May to July makes cloud access a more consequential route for organizations that want to use Llama without managing the underlying deployment themselves.
Second-order effects
- Competing model providers face greater pressure to compete on deployment flexibility, tooling, and cost—not only benchmark performance—as enterprise buyers gain another widely available option.
- Cloud providers serving Llama users can capture more AI workload demand, while Meta’s need to improve developer tools becomes more immediate as adoption broadens.
Third-order effects
- If enterprise and cloud usage continue to grow, open-model distribution could increase buyer leverage by making it easier to compare or switch model access paths rather than relying on a single closed provider.
- The durable contest shifts toward the surrounding infrastructure—hosting, inference economics, developer tooling, and product distribution—rather than model releases alone.
The trend: This is one data point in the shift from frontier-model launches toward competition over who can distribute, host, and operationalize models at scale.