Meta has launched a new open-weight artificial intelligence model, claiming it matches and, in some cases, outperforms competing products from leading Chinese developers.
Gatekeepers News reports that Alexandr Wang, Meta’s chief of AI, announced the launch on Thursday, describing Muse Spark Open as “the most capable US open-weight model to date” in a post on X.
Unlike closed AI models such as Anthropic’s Claude and OpenAI’s ChatGPT, open-weight models allow developers to download their parameters, or weights, and customise them for different applications.
Chinese developers have emerged as major players in the open-weight AI market in recent years, with Alibaba’s Qwen, Moonshot AI’s Kimi K3 and Z.ai’s GLM models regarded as leading competitors.
Meta said its new model was designed to compete with these systems, particularly in advanced programming and complex reasoning tasks.
“Built for coding and agentic work,” Muse Spark Open “excels at long-horizon tasks, complex software engineering, and multi-step reasoning,” the company said in a blog post.
Meta added that “strong safeguards” had been integrated into the model to address potential risks associated with its capabilities.
The company also said Muse Spark Open had been optimised to run on Apple’s advanced M5 Ultra chips, highlighting an unusual collaboration between the two technology giants.
Meta chief executive Mark Zuckerberg said in a post on X that the model “runs great on new Mac Studios thanks to work” with Apple and its chief executive, John Ternus.
Ternus, who succeeded Tim Cook as Apple’s chief in September, also praised the model’s performance on the company’s hardware.
“It’s amazing to run a model of Muse Spark Open’s caliber and scale this efficiently, right from your desk. Mac Studio keeps pushing the boundaries of on-device AI!” he said.
Despite the optimisation, running Muse Spark Open still requires substantial computing resources.
Meta said users could connect two Mac Studios powered by M5 Ultra chips to build what it described as a desktop “supercomputer” capable of running the model.
The launch comes days after Meta introduced two new closed AI models, Wonder and Wonder Full, on Monday. The company claimed the models could compete with the most advanced versions of ChatGPT and Claude.
Meta initially concentrated its AI strategy on open models, particularly its Llama family. However, the company has increasingly shifted towards competing directly with OpenAI and Anthropic in the development of advanced proprietary AI systems.
The company introduced its Spark model in April, marking a departure from its earlier Llama-based approach, which had struggled to keep pace with competitors across parts of the AI industry.
With Muse Spark Open, Meta is seeking to strengthen its position in the open-weight market while challenging the growing influence of Chinese developers in the sector.
