Meta released Muse Ember, its smallest and most efficient AI model to date, specifically engineered to handle the repetitive, high-frequency subtasks that autonomous AI agents execute continuously. Designed as a dedicated execution engine, Ember excels at swift tool calling, API routing, localized coding subtasks, and large-scale document processing without burning heavy compute budgets. Meta aggressively priced the model at an ultra-low $0.10 per million input tokens, making it one of the most cost-effective inference options on the market for enterprise agent loops. Despite its lightweight footprint, Ember retains a robust 1-million-token context window, enabling it to digest extensive logs, codebases, and documents during multi-step background operations. Tech analysts note that the release addresses a critical industry bottleneck reducing the cost of running thousands of background micro-decisions and tool executions required by modern multi-agent systems.

Full report : Tech giant launches its smallest and most cost-efficient model yet, offering a 1-million-token context window at rock-bottom API pricing to power repetitive background agent workflows.