On Tuesday, Meta held her first conference for Developers of him, Llamaco, at its headquarters Menlo Park, California. The company announced the start of a Meta Chatbot application facing the consumer, who will compete with the chatgpt, as well as an API facing the developer to enter LMama models in the Cloud.
Both omissions aim to expand the adoption of Open Llama company models, but this purpose can be secondary to the true defective motive: beating Openai. The ambition of the flaw, with wide strokes, are promoting an open flowering ecosystem of the one that climbs it to “close” the providers of the one as Openai, which ports their models after services.
The application he has a chatbot feels almost like a prejudice to the rumored Openai social network. It has a social food where users can share their conversations, and offers personalized answers based on a user’s meta application activity.
As for API Llama, it is a challenge for Openai’s API business. Llama API is designed to make it easier for developers to build applications associated with Llama models in cloud, using only a single line of code. It eliminates the need to rely on third -party cloud providers to run Llama models, and allows Meta to offer a group of fuller tools to developers.
Meta, like many companies, perceives Opennai as a high rival. Judicial registrations in a case against Meta reveal that the company’s executions were previously obsessed after the Openai GPT-4 beating, which was once a better model. He has long been a substantive strategy for the Meta strategy. In a July 2024 letter, Meta Mark Zuckerberg’s CEO tried to contradict Meta with companies like Openai, writing that “Sale of entry into models is not the business model (meta).”
Some scholars of the one who spoke with Techcrunch before Llamaco were hoping that Meta would issue a competitive model of reasoning like O3-Minini of Openai. The company did not end up doing that. But for Meta, it is not about winning the race of it necessarily.
During a CEO’s CEO conversation with Ali Ghodsi during Llamaco, Zuckerberg said he sees every lab that makes his models open available, including Deepseek and Qwen of Alibaba, as allies in the fight against closed models providers.
“A portion of the value about the open source is that you can mix and match. So if another model, like Deepseek, is better – or if QWen is better in something – then, as developers, you have the best parts of intelligence from different models and produce exactly what you need,” Zuckerberg said. “This is part of the way I think the open source essentially passes in quality the entire closed source (patterns) … (i) feel like a kind of unstoppable force.”
Beyond the overwhelming Openai growth, Meta can also try to push its open patterns to satisfy a regulatory engraving. The EU act it gives special privileges to companies that distribute the systems of “free and open source”. Meta often claims that its Llama models are “open source”, despite disagreements if they meet the necessary criteria.
Despite the reason, Meta seems to be pleased to begin omissions of the one that strengthens the open model ecosystem and limit the growth of Openai-sometimes at the expense of failure to give the front models themselves.
