Meta Platforms Inc. saw its stock edge lower as investors reacted to a notable strategic shift in its artificial intelligence development. The company is moving away from its previous pure open-source model, championed with its Llama family, toward a hybrid approach that blends open-source releases with keeping its most advanced models private.
The new generation of AI models is being developed under the leadership of Alexandr Wang, founder of Scale AI, who now oversees parts of Meta's advanced AI efforts. The company plans to release some versions of these models under open-source licenses, allowing developer modification, while safeguarding its most powerful systems. This dual-track strategy aims to preserve Meta's influence in the open-source ecosystem while protecting core intellectual property from competitors and addressing safety concerns.
The shift comes as Meta faces intensifying competitive pressure. Its Llama 4 model reportedly lagged on several benchmarks, while rivals like OpenAI and Anthropic are preparing next-generation systems. The broader industry trend, seen with players like Alibaba, is also moving toward more closed ecosystems for frontier AI. Meta's hybrid model is an attempt to balance developer access with competitive protection, leveraging its unique strength in consumer reach across platforms like WhatsApp, Facebook, and Instagram to deliver AI at a global scale.
Leadership changes and internal delays have influenced this strategic pivot. Wang's expanded role follows Meta's multibillion-dollar investment in Scale AI. Internally, a key model codenamed "Avocado" slipped from its expected timeline, raising concerns about Meta's ability to keep pace in performance benchmarks. The company does not expect to outperform rivals in every metric but believes it can differentiate in areas that resonate with everyday users.
