At its inaugural LlamaCon AI developer convention on Tuesday, Meta introduced an API for its Llama collection of AI fashions: the Llama API.
Available in restricted preview, the Llama API lets builders discover and experiment with merchandise powered by totally different Llama fashions, per Meta. Paired with Meta’s SDKs, it permits builders to construct Llama-driven companies, instruments and functions. Meta didn’t instantly share the API’s pricing with TechCrunch.
The rollout of the API comes as Meta appears to take care of a lead within the fiercely aggressive open mannequin area. While Llama fashions have racked up greater than a billion downloads to this point, based on Meta, rivals comparable to DeepSeek and Alibaba’s Qwen threaten to upend Meta’s efforts to determine a far-reaching ecosystem with Llama.
The Llama API gives instruments to fine-tune and consider the efficiency of Llama fashions, beginning with Llama 3.3 8B. Customers can generate knowledge, practice on it, after which use Meta’s analysis suite within the Llama API to check the standard of their customized mannequin.
Meta stated it gained’t use Llama API buyer knowledge to coach the corporate’s personal fashions, and that fashions constructed utilizing the Llama API will be transferred to a different host.
For devs constructing on prime of Meta’s lately launched Llama 4 fashions particularly, the Llama API gives model-serving choices through partnerships with Cerebras and Groq. These “early experimental” choices are “obtainable by request” to assist builders prototype their AI apps, Meta stated.
“By merely choosing the Cerebras or Groq mannequin names within the API, builders can […] get pleasure from a streamlined expertise with all utilization tracked in a single location,” wrote Meta in a weblog put up supplied to TechCrunch. “[W]e sit up for increasing partnerships with extra suppliers to deliver much more choices to construct on prime of Llama.”
Meta stated it’s going to develop entry to the Llama API “within the coming weeks and months.”
