Keynote: The Llama Ecosystem: Past, Present and Future
summary
Llama begins as a scrappy Paris research project and scales to 7B, 13B, 70B models with an instruction-tuned Llama 2 and Code Llama checkpoints (7, 13 and 34B) fine-tuned for coding. The ecosystem shows concrete traction: about 9,400 derivative models, ~32 million downloads in 30 days, and over 8,000 GitHub projects. Hardware and cloud partners (AMD, Qualcomm, Nvidia, Intel, AWS, Google, Cloudflare) enable edge and CDN inference, and community-tuned variants improve leaderboard scores by roughly 10%.
AI-written summary · about
chapters (8)
- 0:00Llama ecosystem overview
- 1:46Models as computing platforms
- 2:28Origins and growth of Llama
- 4:12Llama's impact
- 6:00Community contributions
- 7:47Open LLM leaderboard
- 8:49Full-stack partnerships
- 10:13Future directions
related talks
17mThe State of the Llama Ecosystem
38mMeta's Roadmap for Full Stack AI
31mBuilding LLaMA: Meta's Director of GenAI
01h 07mKeynote + Conversation
11mIntroducing Llama API: A Seamless Way to Revolutionize Your Workflow
48mBringing Llama 3 to Life
12mUnpack the Latest Llama 4 Herd
22mThe Disruptive Potential of On-Device LLMs