On Tuesday, Meta held its first-ever AI developer convention, LlamaCon, at its Menlo Park, California headquarters. The corporate introduced the launch of a consumer-facing Meta AI chatbot app, which is able to compete with ChatGPT, in addition to a developer-facing API for accessing Llama fashions within the cloud.
Each releases purpose to develop adoption of the corporate’s open Llama AI fashions, however that objective could also be secondary to Meta’s true motive: beating OpenAI. Meta’s AI ambition, in broad strokes, is fueling a thriving open AI ecosystem that sticks it to “closed” AI suppliers like OpenAI, which gate their fashions behind companies.
Meta’s AI chatbot app feels virtually like a preemption of OpenAI’s rumored social community. It has a social feed the place customers can share their AI chats, and presents customized responses based mostly on a consumer’s Meta app exercise.
As for the Llama API, it’s a problem to OpenAI’s API enterprise. The Llama API is designed to make it less complicated for builders to construct apps that hook up with Llama fashions within the cloud, utilizing only a single line of code. It eliminates the necessity to depend on third-party cloud suppliers to run Llama fashions, and permits Meta to supply a fuller array of instruments for AI builders.
Meta, like many AI corporations, perceives OpenAI to be a prime rival. Courtroom filings in a case in opposition to Meta reveal that the corporate’s execs beforehand obsessed over beating OpenAI’s GPT-4, which was as soon as a state-of-the-art mannequin. Undercutting proprietary AI mannequin suppliers like OpenAI has lengthy been core to Meta’s AI technique. In a July 2024 letter, Meta CEO Mark Zuckerberg sought to distinction Meta with corporations like OpenAI, writing that “promoting entry to AI fashions isn’t [Meta’s] enterprise mannequin.”
A number of AI researchers who spoke with TechCrunch forward of LlamaCon have been hoping Meta would launch a aggressive AI reasoning mannequin like OpenAI’s o3-mini. The corporate didn’t find yourself doing so. However for Meta, it’s not about successful the AI race essentially.
Throughout an onstage dialog with Databricks CEO Ali Ghodsi throughout LlamaCon, Zuckerberg mentioned he sees any AI lab that makes its fashions overtly obtainable, together with DeepSeek and Alibaba’s Qwen, as allies within the battle in opposition to closed mannequin suppliers.
“A part of the worth round open supply is that you may combine and match. So if one other mannequin, like DeepSeek, is best — or if Qwen is best at one thing — then, as builders, you may have the flexibility to take the most effective components of the intelligence from completely different fashions and produce precisely what you want,” mentioned Zuckerberg. “That is a part of how I feel open supply mainly passes in high quality all of the closed supply [models] … [I]t appears like form of an unstoppable power.”
Past stunting OpenAI’s progress, Meta might also be making an attempt to push its open fashions to fulfill a regulatory carveout. The EU AI Act grants particular privileges to corporations that distribute “free and open supply” AI programs. Meta usually claims its Llama fashions are “open supply,” regardless of disagreement on whether or not they meet the required standards.
Whatever the motive, Meta appears content material to kick off AI launches that strengthen the open mannequin ecosystem and restrict OpenAI’s progress — generally on the expense of failing to ship cutting-edge fashions itself.
