# Meta unveils Muse Spark AI model with paid developer tier, marking a sharp pivot from open-source roots > Source: > Published: 2026-07-09 14:11:52+00:00 # Meta unveils Muse Spark AI model with paid developer tier, marking a sharp pivot from open-source roots The company's first proprietary AI model signals a new revenue strategy that could reshape how Big Tech monetizes artificial intelligence, with ripple effects across crypto-adjacent AI token markets. Meta just did something it spent years telling everyone it wouldn’t do. The company launched Muse Spark, a closed, proprietary AI model, effectively reversing its long-standing commitment to open-weight AI development. The move comes with a planned paid API tier for developers, a first for a company that built its AI reputation on giving things away for free. ## From open to closed, and free to paid Muse Spark launched on April 8, 2026, as the first model from Meta’s new Muse series, built by Meta Superintelligence Labs. It powers the Meta AI app and website, with a focus on visual reasoning, tool usage, and agentic capabilities. The model represents a clean break from Meta’s Llama series, the open-weight models released throughout 2024 and 2025 that anyone could download, modify, and deploy. Meta provided a private API preview to select partners after launch, with plans to eventually open a public developer API with paid access tiers. By June 3, the company had delayed the public API rollout multiple times without committing to a firm release date. No specific pricing details or tier structures have been disclosed. ## Why the pivot matters beyond Silicon Valley Performance-wise, the model shows genuine strength in multimodal reasoning benchmarks, meaning tasks that require understanding both text and images simultaneously. It also performs well in tool usage scenarios. But it has notable weaknesses in coding tasks, which is a gap its competitors have been aggressively closing. ## The crypto-AI connection investors shouldn’t ignore There’s no direct crypto component to the Muse Spark announcement. Meta hasn’t mentioned tokens, blockchain integration, or decentralized inference networks. Meta’s repeated delays in launching even its centralized API highlight just how difficult it is to offer reliable, scalable model access. If a company with Meta’s engineering resources struggles with API rollout timelines, decentralized alternatives face exponentially harder execution challenges. **Disclosure:** This article was edited by Editorial Team. For more information on how we create and review content, see our [Editorial Policy](https://cryptobriefing.com/editorial-policy/).