Meta launched its first-ever LlamaCon event with several announcements aimed at expanding the Llama ecosystem, which has surpassed one billion downloads since its launch two years ago. The company introduced new developer tools, security features, and grant recipients focused on advancing open-source AI development.
New Llama API Platform for Developers
Meta has announced the Llama API, available as a limited, free preview, which combines the features of closed model APIs with the open-source flexibility of the Llama model. The platform offers one-click API key creation and interactive playgrounds for exploring Llama models, including the recently released Llama 4 Scout and Llama 4 Maverick.
The API includes a lightweight SDK in both Python and TypeScript, offering compatibility with the OpenAI SDK to facilitate the easy conversion of existing applications. Meta emphasizes that user data privacy is a top priority, stating that they do not use prompts or model responses to train their AI models.
The preview release includes tools for fine-tuning and evaluation of custom versions of the new Llama 3.3 8B model. Meta plans a broader rollout in the coming months, with interested developers able to apply for early access.
Accelerated Inference and Partner Integrations
Meta announced collaborations with Cerebras and Groq to provide faster inference speeds through the Llama API. Developers can access Llama 4 models powered by these providers via the API for experimentation and prototyping.
The company is also expanding its Llama Stack distributions through integrations with NVIDIA NeMo microservices and partnerships with IBM, Red Hat, and Dell Technologies, with the aim of establishing Llama Stack as an industry standard for enterprise AI deployments.
Enhanced Protection Tools
In response to security concerns surrounding increasingly capable AI models, Meta has released new protection tools, including Llama Guard 4, Llama Firewall, and Llama Prompt Guard 2. The company also announces updates to CyberSecEval 4 for evaluating AI systems in security operations, along with the Llama Defenders Program for select partners.
Llama Impact Grants
Meta announced the 10 international recipients of its second Llama Impact Grants, awarding over $1.5 million to support companies, startups, and universities utilizing Llama for social impact.
Recipients include E.E.R.S. from the U.S., which improves public access to civic services, and Doses AI in the UK, which enhances pharmacy operations with real-time error detection. Other notable winners include Solo Tech, providing offline AI support in rural U.S. communities, and FoondaMate, a multilingual study tool serving millions of students across Africa.
