Meta’s Llama 4 model behind schedule, underperforming: Report
Meta has bet a ton — some would even say the future of the company — on AI.
It has poured massive resources into building its AI infrastructure to train and run its Llama 3 AI model, including city-sized data centers and possibly a $200 billion data center campus.
But everything depends on its next model: Llama 4. And according to new reporting from The Information, things are not going according to plan. Development has been subject to several delays, and the release of DeepSeek’s breakthrough fast and cheap models caused Meta to make significant changes to its development plans.
The company is trying to get to 1 billion AI users and then monetize them, but it currently gives Llama 3 away for free to developers. Per the report, Meta is considering selling businesses API access to a hosted version of the model, which it currently does not offer, unlike startup competitors OpenAI and Anthropic.
The rest of the AI industry is also trying to figure out how to justify the hundreds of billions in capex spending planned for this year to its investors.
But everything depends on its next model: Llama 4. And according to new reporting from The Information, things are not going according to plan. Development has been subject to several delays, and the release of DeepSeek’s breakthrough fast and cheap models caused Meta to make significant changes to its development plans.
The company is trying to get to 1 billion AI users and then monetize them, but it currently gives Llama 3 away for free to developers. Per the report, Meta is considering selling businesses API access to a hosted version of the model, which it currently does not offer, unlike startup competitors OpenAI and Anthropic.
The rest of the AI industry is also trying to figure out how to justify the hundreds of billions in capex spending planned for this year to its investors.