Meta releases Muse Spark Open, a new ‘open-weight’ AI model
Meta has launched Muse Spark Open, an open-weight AI model it says can match or outperform leading Chinese models. The release matters because Chinese developers have led the open-weight market, while Meta is trying to strengthen its position in AI.
Meta says the model is designed for coding and agentic work, including complex software engineering and multi-step reasoning, and has safeguards built in. It is optimised for Apple’s M5 Ultra chips, but requires substantial computing power: Meta says users can run it across two Mac Studios to create a desktop “supercomputer”. The launch follows Meta’s announcement of a new closed model intended to compete with advanced offerings from OpenAI and Anthropic.
- Meta has released its new open-weight model, Muse Spark Open.
- The company says it can compete with leading Chinese models.
- Running it may require two Mac Studios with M5 Ultra chips.
New here? Start with this
Open-weight AI models make their trained parameters available for others to download and use, giving developers more control over how they adapt and run them. They are different from closed models, which are accessed through a company’s service and whose underlying parameters are not released.
Meta is the company behind Facebook, Instagram and WhatsApp, and it is investing in AI to compete with firms such as OpenAI and Anthropic. Chinese developers have been prominent in the open-weight market, so Meta’s release is also a move to build its presence there.
The model is aimed at tasks such as writing code and carrying out work in several steps. Running it may require powerful hardware, which affects who can use it and at what cost; Meta says it can run across two Mac Studios.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Releasing model weights can widen access to capable AI, letting developers inspect, adapt and run the model themselves, and potentially build useful tools without depending on a single provider. Meta’s move may also strengthen competition in a market where Chinese developers have led open-weight releases, while its stated safeguards could help address risks.
The case against
Open weights can make it harder to control how a powerful model is modified or used, so safeguards at launch may not prevent harmful downstream applications. The substantial computing requirements may also limit who can actually use the model, while Meta’s performance and safety claims need independent scrutiny before developers can judge its practical value.