Meta has officially launched 'Muse Spark,' a compact yet high-performance AI model designed for complex reasoning in science, math, and health. Developed by the newly formed Meta Superintelligence Labs under the leadership of Alexandr Wang, the model features a 'Contemplating' mode intended to rival the reasoning capabilities of OpenAI’s 'Thinking' series and Google’s Gemini Deep Think. The release signals Meta's successful ground-up rebuild of its AI stack after a multi-billion dollar investment cycle.
AI is transitioning from a productivity tool into an autonomous 'intelligence layer.' Major players are reporting a massive surge in 'agentic' workflows, where multi-agent systems execute end-to-end business processes—such as lead scoring and CRM updates—without human intervention. OpenAI recently revealed that enterprise demand now accounts for over 40% of its total revenue, driven by this shift.
Meta's projected capital expenditure for 2026 is expected to peak at $135 billion—nearly double its 2025 spending—as the company builds the infrastructure required for 'personal superintelligence.'
Which new Meta AI model, code-named 'Avocado' during development, was released today to compete with Google and OpenAI in reasoning tasks?