IRELAND-I-HAD-SUMMIT
Meta has launched a new open-weighted artificial intelligence model, targeting the rapidly growing open model market. Model performance, computational cost, and data security will be key performance indicators for future competition.

US technology company Meta has launched a new open-weight artificial intelligence (AI) model as it seeks to compete with Chinese developers in the rapidly growing open-model market. Meta said the model can go head-to-head with products from leading Chinese developers in some capability tests, highlighting how global AI competition is expanding beyond closed services to a model ecosystem that can be downloaded and customized.

According to Central News Agency, Meta released the new model, Muse Spark Open, on Oct. 15. Meta AI chief Alexandr Wang described it as one of the company’s most powerful open-weight models developed in the United States to date, saying it was designed for coding, complex software engineering, long-running tasks and multistep reasoning.

Unlike conventional closed AI services, open-weight models make their model weights available to developers, allowing them to deploy, modify and customize the systems for their own needs. This approach gives businesses and researchers greater flexibility, while making model performance, computing costs and security management key areas of competition.

In recent years, models developed by Chinese companies, including Qwen, Kimi and GLM, have attracted attention in the open-weight market. Meta’s launch reflects efforts by US technology companies to strengthen the competitiveness of open models and respond to the challenges posed by Chinese products in pricing, customization and code generation.

Meta also said Muse Spark Open has been optimized for Apple’s M5 Ultra chip. Reports indicate that running the model on a desktop system could require two Mac Studio computers equipped with M5 Ultra chips, underscoring the substantial hardware and computing demands of large models.

For the global technology industry, the development of open models could give businesses more options for adopting AI. However, model performance, deployment costs, data security and safeguards remain factors that organizations must assess before adoption. Meta’s claimed performance advantages also require validation through further independent testing and real-world applications.