Hajra Naz
At its Menlo Park offices in California, Meta hosted LlamaCon, its first AI development event, on Tuesday. Meta made two significant AI announcements at the event:
-
A new Meta AI chatbot app aimed at everyday users
-
A developer-friendly API that allows direct access to its Llama models in the cloud
These tools are designed to broaden the use of Meta’s Llama models and deepen its footprint in the growing AI landscape.
Meta AI App
OpenAI’s ChatGPT is positioned to compete with the new Meta AI chatbot software. Users may post AI-generated discussions in the app’s social feed, which also provides personalized responses based on user activity across all of Meta’s platforms.
This launch seems strategically timed—possibly to get ahead of OpenAI’s rumored AI-powered social network. Meta appears to be trying to blend chat, personalization, and social sharing in a way that’s difficult for OpenAI to match.
Llama API: Targeting Developers and OpenAI’s Core Business
With just one line of code, developers may connect to Llama models thanks to Meta’s introduction of the Llama API. This tool removes the need for external cloud providers and gives developers a direct way to build apps powered by Llama AI.
The API is a direct shot at OpenAI’s API services, offering developers another path with fewer restrictions and potentially lower costs.
Undercutting Closed Models: A Core Part of Meta’s AI Strategy
Meta’s broader strategy is becoming clear: challenge companies like OpenAI that keep their models locked behind paywalls. For years, Meta executives have made it clear they see OpenAI as a major competitor. Internal court documents even revealed that Meta’s leadership was fixated on outperforming OpenAI’s GPT-4.
In a July 2024 letter, CEO Mark Zuckerberg stated:
“Selling access to AI models isn’t [Meta’s] business model.”
This position is part of Meta’s push to set itself apart from rivals who charge for model access.
No New Reasoning Model—But a Message of Collaboration
Some AI researchers were expecting Meta to unveil a new reasoning model to rival OpenAI’s o3-mini ahead of LlamaCon. That didn’t happen. But during a talk with Databricks CEO Ali Ghodsi, Zuckerberg emphasized that Meta sees open labs like DeepSeek and Alibaba’s Qwen as partners rather than threats.
“If DeepSeek is better at something, or if Qwen is, developers can mix and match,” Zuckerberg said. “This is how open source eventually overtakes closed models. It’s becoming an unstoppable force.”
Regulatory Incentives: Another Reason for Open Models?
While competing with OpenAI is a key driver, Meta may also have its eye on regulatory benefits. Providers of “free and open source” models are given preferential consideration under the EU AI Act. Meta often claims its Llama models fall into that category, though there’s still debate over whether they truly qualify.
Strengthening the Open AI Ecosystem—Even Without Flashy Models
In the end, Meta might not be focused on producing the most powerful model on the market. Instead, it’s betting on a bigger picture strategy:
-
Encourage openness in the AI world
-
Challenge dominant closed systems
-
Grow its influence by offering useful tools rather than a monetizing model access
Whether it’s about regulatory compliance or competition, Meta seems willing to trade short-term wins for long-term influence in shaping the AI ecosystem.
