Meta Launches Llama API for AI Developers (Preview)

Meta Releases Llama API in Limited Preview

Meta announced the Llama API at its inaugural LlamaCon AI developer conference. This API grants developers access to Meta's Llama series of AI models, enabling them to experiment and build innovative AI-powered products.

Currently in limited preview, the Llama API, combined with Meta's SDKs, empowers developers to create Llama-driven services, tools, and applications. Developers can apply to join the waitlist for access.

Staying Competitive in the Open-Source AI Landscape

With over a billion downloads of Llama models to date, Meta aims to maintain its leadership in the competitive open-source AI model arena. The Llama API is a key component of this strategy, facing competition from rivals like DeepSeek and Alibaba's Qwen.

Fine-Tuning and Evaluating Llama Models

The Llama API provides tools for fine-tuning and evaluating Llama model performance, starting with Llama 3.3 8B. Developers can generate data, train their models, and utilize Meta's evaluation suite within the API to assess the quality of their custom models.

Meta assures developers that data used with the Llama API will not be used to train Meta's own models. Furthermore, models built using the API are transferable to other hosting platforms.

Model Serving with Cerebras and Groq

For developers working with the recently released Llama 4 models, the Llama API offers model-serving options through partnerships with Cerebras and Groq. These early experimental options are available by request, enabling developers to prototype their AI applications.

By simply selecting the Cerebras or Groq model names in the API, developers can enjoy a streamlined experience with all usage tracked in one location. We look forward to expanding partnerships with additional providers to bring even more options to build on top of Llama.

Meta plans to expand access to the Llama API in the coming weeks and months. More information is available on the Llama API waitlist page.