Meta abandons open-source Llama for proprietary Muse Spark
Meta has shifted its focus from the open-source Llama AI models to a new proprietary model called Muse Spark, developed by its Superintelligence Labs team. While existing Llama models will remain available, future development is expected to be limited, leaving users without a clear migration path to the new cloud-only Muse Spark. This move marks a significant departure from Meta's earlier support for open-source AI and presents challenges for developers relying on Llama for self-hosted or customizable solutions.
Opening excerpt (first ~120 words) tap to expand
Meta’s decided its new proprietary AI LLM, Muse Spark, would be much more profitable than the open-source Llama. What are Llama users to do? Back in 2023, Meta released Llama 2.0 and proclaimed it to be open source. Even if Llama wasn’t actually open source, it sounded good to developers. Indeed, in October 2024, Meta founder and CEO Mark Zuckerberg proclaimed, “Open Source AI is the Path Forward.” In March of last year, the company issued a press release celebrating 1 billion downloads of Llama. That was then. This is now. Meta’s pivot to proprietary AI For all practical purposes, Meta has abandoned developing Llama in favor of its new proprietary program, Muse Spark, announced this month.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at The New Stack.
