Meta introduced a new open-weight artificial intelligence model called Muse Spark Open on Thursday, claiming it matches or surpasses comparable products created by leading Chinese developers.
Posting on X, Alexandr Wang, chief of AI at Meta, described Muse Spark Open as the most capable US open-weight model produced to date.
Over the last few years, Chinese tech firms have established dominance in the global open-weight AI sector, with Alibaba’s Qwen, Moonshot AI’s Kimi K3, and Z.ai’s GLM models widely regarded as international frontrunners, reports Guardian.
However, Meta asserts that its latest release now operates at an equivalent level.
In contrast to closed models such as OpenAI’s ChatGPT or Anthropic’s Claude, open-weight models permit developers to download and modify their core parameters, known as weights and biases.
In a blogpost, Meta stated that Muse Spark Open was built specifically for coding and agentic work, highlighting its strength in handling complex software engineering, multi-step reasoning, and long-horizon tasks.
The company added that strong safeguards have been integrated into the new architecture.
Due to its heavy computational requirements, the model has been optimised for Apple’s advanced M5 Ultra chips, marking an unusual partnership between the two technology companies.
According to Meta, users can construct a personal desktop “supercomputer” by running the model on two Mac Studio devices equipped with M5 Ultra processors.
The launch comes shortly after Meta announced its newest closed model, which it claimed rivalled the most advanced versions of Claude and ChatGPT.
Although Meta’s artificial intelligence efforts were initially focused on open models, Mark Zuckerberg, chief executive of Meta, has more recently pushed the firm to compete directly with OpenAI and Anthropic.
