LlamaCon has officially started. This first-ever event brings developers from all over the world who share a common passion: building with Llama. It’s been just over two years since Llama launched. In that time, it has reached over one billion downloads and become a leader in the open-source AI ecosystem. Developers, startups, governments, and businesses are all using Llama to drive innovation, improve efficiency, and solve tough problems.
Of course, no event would be complete without sharing new open-source tools. These tools will help developers explore Llama models, build faster, and use the latest resources to protect and defend their work.
Here’s what’s new today and how you can start using these LlamaCon 2025 releases.
Here is Everythings You Need to Know About LlamaCon 2025
Try Llama API (Preview): Powerful Like Closed Models, Flexible Like Open Source
Meta has made it easier for developers to start building with Llama. They offer full control over models and weights—without locking users into a specific API. They recently launched the Llama API as a limited free preview. This new developer platform helps users build Llama-powered applications quickly.
With Llama API, developers can generate an API key in just one click. They can explore different Llama models—like Llama 4 Scout and Llama 4 Maverick—using interactive playgrounds. When it’s time to build, developers can use lightweight SDKs for Python or TypeScript. The Llama API also supports the OpenAI SDK, making it easy to migrate or extend existing applications.
Meta knows developers need tools to customize and run models more efficiently. As part of this release, they’ve added fine-tuning and evaluation tools to the new Llama API. Developers can now fine-tune their own custom versions of the new Llama 3.3 8B model.
These tools help lower costs while improving speed and accuracy. Developers can generate training data, train their models, and test them easily using Meta’s built-in evaluation suite. This makes it simple to measure model performance with real data instead of guessing.
Meta also prioritizes security and privacy. They don’t use your prompts or model responses to train their AI. When developers create models using the Llama API, they can take those models anywhere—they aren’t locked into Meta’s servers.
Right now, Meta is offering these features to a small group of users, with plans to expand access in the coming weeks. This update opens the door for developers to build powerful custom models for all kinds of use cases.
This preview is just the beginning for the Llama API. Meta wants more developers to start building and sharing feedback to shape future improvements. Anyone interested in joining the free preview can apply by filling out a short form.
Fast Inference with Llama API
Meta has teamed up with Cerebras and Groq to help developers run Llama models faster. Developers can now request early experimental access to Llama 4 models powered by Cerebras and Groq. This allows them to test and prototype quickly before scaling with a preferred vendor.
To use these faster models, developers simply select the Cerebras or Groq model name in the Llama API. The API tracks all usage in one place, making the experience smooth and simple. This collaboration shows Meta’s commitment to giving developers more flexibility, and they plan to add more partners soon to expand these options.
New Llama Stack Integrations
Meta listened to developers who wanted easier ways to deploy apps across various platforms. Last year, they worked with partners to create Llama Stack distributions for enterprise users. This year, they are building on that effort.
They recently integrated Llama Stack with NVIDIA NeMo microservices. They’re also working closely with companies like IBM, Red Hat, and Dell Technologies to launch more integrations soon. Meta and its partners aim to make Llama Stack the go-to solution for deploying AI applications at an enterprise level.
New Llama Security Tools for Open Source
Meta continues to support developers with tools that make AI applications more secure. Today, they launched new protection tools for the open source community: Llama Guard 4, LlamaFirewall, and Llama Prompt Guard 2.
They also shared updates on CyberSecEval 4, which helps organizations test how well AI systems perform in security roles. Additionally, they announced the Llama Defenders Program for select partners. As AI models grow more powerful, Meta sees these tools as key to building safer and stronger software.
Meet the Llama Impact Grant Recipients
The Llama Impact Grants help promote innovation and create economic opportunities through open-source AI. Today, Meta is excited to announce the 10 international recipients of the second Llama Impact Grants. These grants, totaling over $1.5 million USD, support companies, startups, and universities using Llama to drive meaningful change.
This year’s recipients include a wide range of solutions. For example, E.E.R.S. from the US uses a Llama-enabled chatbot to improve public access to civic services. In the UK, Doses AI uses Llama to improve pharmacy operations with real-time error detection. Other notable winners include Solo Tech, which provides offline AI support to rural US communities, and FoondaMate, a multilingual study tool that helps millions of students in Africa and beyond.
The Future is Open Source
Meta leads the way in open-source AI, with the Llama ecosystem at the heart of this revolution. With Llama, developers and businesses have the freedom to create whatever they need—without being tied to proprietary systems or long contracts. This flexibility, along with portability and accessibility, makes Llama the top choice for those wanting to tap into AI’s power.
Meta is committed to being a long-term partner for developers and businesses, offering a smooth transition from closed models. Llama is affordable, easy to use, and helps more people access AI, no matter their technical skills or hardware. Meta believes in AI’s ability to transform industries and improve lives, which is why they’re excited to support Llama’s growth and development. They can’t wait to see what users will create next.
Conclusion:
LlamaCon 2025 has truly set the stage for a new era in AI development. With exciting new tools, models, and updates unveiled, it’s clear that Llama continues to lead the way in the open-source AI ecosystem. Developers now have even more powerful resources at their fingertips to create innovative solutions and push the boundaries of what’s possible. Whether you’re just getting started or looking to expand your Llama projects, the announcements from this year’s event provide everything you need to take your development to the next level. Stay tuned for more updates as Llama continues to evolve and shape the future of Artificial Intelligence.