Meta Launches Muse Spark, Its First Proprietary AI Model With No Open Source

by Admin
Meta Launches Muse Spark

Meta has introduced Muse Spark, a proprietary AI model developed by its Superintelligence Labs team. Unlike the Llama models, Muse Spark’s weights are not publicly accessible.

Currently, access is limited to Meta’s AI portal at meta.ai, with API access granted by invitation only. According to a blog post announcing the release, Meta describes Muse Spark as “the first step on our scaling ladder and the first product of a ground-up overhaul of our AI efforts.”

What Meta’s Muse Spark Model Is

Meta describes the model as a natively multimodal reasoning system equipped with tool use, visual chain of thought, and multi-agent coordination features. The company has not revealed specifics about its underlying architecture or the number of parameters.

It has also introduced a “Contemplating mode” that runs multiple reasoning agents in parallel, aiming to compete with Gemini Deep Think and GPT Pro. This mode is not available at launch and is being gradually rolled out through meta.ai.

Muse Spark Performance vs. Other AI Models

Meta states that Muse Spark performs on par with or exceeds the benchmarks set by leading models from OpenAI, Anthropic, and Google. The company also claims that the model required significantly less compute to train compared to Llama 4 while achieving similar performance levels.

Meta has published its testing methodology alongside the benchmark results. These figures are Meta’s own representations. Earlier this year, Meta faced criticism over its benchmarking practices for the Llama 4 release.

Open-Source Pivot And Meta’s Superintelligence Labs Roadmap

The closed release signals a notable change from what Meta CEO Mark Zuckerberg described in a 2024 post titled “Open Source AI is the Path Forward.” In that post, he argued that open source AI was the best way to achieve broad economic benefits and that releasing Llama did not threaten Meta’s business model.

Muse Spark follows Meta’s Llama API inference service, launched alongside Llama 4. This move marked a step toward monetizing access to these models. Zuckerberg also shared in a Threads post that Meta still intends to release open-source models in the future alongside its proprietary options.

Muse Spark was developed by Meta Superintelligence Labs, a team formed after Meta ceased work on its largest Llama 4 variant, called Behemoth, which was planned to have 2 trillion parameters.

Meta has not revealed the size of Muse Spark or the other Muse variants that are already in development. It has been confirmed that additional models in the Muse family are in development, but no release dates have been announced.

Related Posts

Leave a Comment