Meta has reportedly delayed the release of the largest version of its open-source Llama 4 artificial intelligence (AI) model, known as ‘Behemoth,’ from summer to fall at the earliest. The multimodal model, which was already postponed from April, is not showing ‘significant’ enough improvements to warrant a June release. This marks the first setback for Meta in releasing its Llama family of large language models, which have been praised for their rapid development. As an open-source model, Llama provides developers in smaller companies, nonprofit communities, and academia with a generally free AI tool, contrasting with proprietary models from OpenAI, Google, and Amazon. However, smaller firms still require assistance to implement these models since Meta does not offer deployment services. Meta uses Llama to enhance its own social media tools, allowing CEO Mark Zuckerberg to maintain control over his AI initiatives. The delay raises questions about whether Behemoth demonstrates sufficient advancements to justify its public launch. Meta’s leadership is reportedly growing frustrated with the progress of its Llama 4 team, leading to potential leadership changes within its AI product group. Despite promoting Behemoth as a superior system internally, training challenges have hindered its effectiveness. Meta has not yet responded to inquiries regarding this matter. Meanwhile, other companies like OpenAI have also faced delays in their AI model releases.
— new from PYMNTS.com
