Tech
Meta vs OpenAI: How LlamaCon Signals a Bold New AI Rivalry
Meta vs OpenAI
At its first-ever AI developer conference, LlamaCon, held Tuesday at its Menlo Park, California offices, Meta announced the release of a developer-facing API for cloud access to Llama models and a consumer-facing Meta AI chatbot software that would compete with ChatGPT.
While increasing the use of the company’s open Llama AI models is the primary objective of both releases, Meta’s real motivation may be to outperform OpenAI. In general, Meta’s AI aspirations are supporting a flourishing open AI ecosystem that defers to “closed” AI providers like OpenAI, who encrypt their models.
It nearly seems as though Meta’s AI chatbot software preempts OpenAI’s speculated social network. It provides tailored replies based on a user’s Meta app behavior and has a social feed where users may share their AI discussions.
The Llama API presents a problem for OpenAI’s API business. With only one line of code, developers can now more easily create apps that link to Llama models in the cloud thanks to the Llama API. It enables Meta to provide a wider range of tools for AI developers and removes the need to rely on other cloud providers to run Llama models.
Like many other AI firms, Meta considers OpenAI to be a major competitor. According to court documents in a complaint against Meta, the company’s executives were originally fixated with defeating OpenAI’s GPT-4, a once-state-of-the-art model. A key component of Meta’s AI strategy has long been undermining proprietary AI model suppliers such as OpenAI. Mark Zuckerberg, the CEO of Meta, attempted to draw a comparison between Meta and businesses such as OpenAI in a letter dated July 2024, stating that “selling access to AI models isn’t [Meta’s] business model.”
Before LlamaCon, a number of AI experts told TechCrunch that they hoped Meta will release a competitive AI reasoning model similar to OpenAI’s o3-mini. The business ultimately chose not to. However, Meta doesn’t necessarily care about winning the AI race.
Zuckerberg stated that he views any AI lab that makes its models publicly available, such as DeepSeek and Alibaba’s Qwen, as friends in the struggle against closed model suppliers during an onstage discussion with Databricks CEO Ali Ghodsi during LlamaCon.
The versatility of open source contributes to its value. According to Zuckerberg, developers may combine the finest features of many models to create precisely what they want, even if another model, such as DeepSeek, is superior or Qwen excels at something. This is an example of how, in my opinion, open source models essentially outperform all closed source models in terms of quality. [I]t seems like an unstoppable power.
In addition to impeding OpenAI’s expansion, Meta could be attempting to promote its open models in order to meet legal requirements. Businesses who release “free and open source” AI technologies are given preferential treatment under the EU AI Act. Although there is debate about whether Meta’s Llama models satisfy the requirements, the company frequently asserts that they are “open source.”
For whatever reason, Meta is happy to initiate AI launches that fortify the open model ecosystem and restrain the expansion of OpenAI, often at the price of its own inability to produce innovative models.