Our Smallest, Most Efficient Model: Muse Ember (research.meta.ai)

Today marks the launch of Muse Ember, touted as the smallest and most efficient AI model yet, tailored for complex reasoning tasks such as tool calls, coding subtasks, and document parsing. Standing out in head-to-head comparisons, Muse Ember delivers superior performance at a fraction of the cost, achieving optimal scores against competitors like Claude Haiku 5.5 and GPT-6 Luna while also topping the efficiency rankings on the Artificial Analysis leaderboard. Specifically, it excels in various key evaluations, including automation workflows and coding tasks, underscoring its advanced capabilities with results that far surpass alternatives. This release is significant for the AI/ML community as it not only demonstrates a remarkable stride in cost-effectiveness but also highlights a growing trend toward developing compact yet powerful models capable of handling sophisticated reasoning processes. Priced affordably at $0.10 per million input tokens and $0.50 per million output tokens, Muse Ember is immediately available via the Meta Model API, allowing developers and businesses to harness its capabilities for repetitive tasks such as contract reviews and code generation, ultimately fostering increased productivity and efficiency in AI-driven operations.