At the first Llamacon AI developer conference held on Tuesday, Meta announced the API for the Llama API, the Llama series of AI models.
The Llama API is available in a limited preview, allowing developers to explore and experiment with products with different Llama models for each meta. In combination with Meta’s SDK, developers can build Llama-driven services, tools and applications. Meta did not immediately share API pricing with TechCrunch.
The API deployment is as Meta tries to maintain his lead in a fiercely competitive open model space. The Llama model has so far won over 1 billion downloads, but according to Meta, rivals such as Deepseek and Alibaba Qwen have threatened to overturn Meta’s efforts to establish a broad ecosystem with the Lama.
The Llama API provides tools for tweaking and evaluating the performance of Llama Models starting with Llama 3.3 8b. Customers can generate data and train on it, then use Meta’s assessment suite in the Llama API to test the quality of their custom models.
Meta said it would not train the company’s own models using customer data from the Llama API, and said that models built using the Llama API can be transferred to another host.
For developers built on Meta’s recently released Llama 4 models, the Llama API in particular offers model serving options through a partnership between Cerebras and GROQ. These “early experiment” options are “available on request” to help developers prototype AI apps, Meta said.
“Simply select a Celebras or GROQ model name in the API and developers can enjoy a streamlined experience with all usages tracked in one location,” wrote the meta in a blog post provided to TechCrunch. “(w) I look forward to expanding my partnership with additional providers and bringing more options on top of the llama.”
Meta said it will expand access to the Llama API “in the coming weeks and months.”