Muse Spark 1.1 has achieved a score of 69 on the Artificial Analysis Coding Agent Index in the Opencode harness, positioning it just below GPT-5.5 and above Claude Opus 4.8 in comparative performance metrics. Priced at approximately $1.4 per task, it offers strong cost-efficiency, though with a longer time required per task. This development reflects Meta’s commitment to open science methodologies in advancing AI coding agents, an effort actively supported by its leadership team.
Finkd: Finkd is the X handle for Mark Zuckerberg, CEO of Meta, who provides overarching leadership for the company’s AI research and product directions. He is directly credited alongside other key figures for the Muse Spark 1.1 achievement. This recognition reflects executive oversight of Meta’s AI coding agent projects.
AI at Meta: AI at Meta is Meta’s research organization dedicated to advancing open AI models and agentic systems through community collaboration. It focuses on pushing technical boundaries in areas such as coding and reasoning capabilities. The team played a central role in the development and evaluation of Muse Spark 1.1.
Alexandr Wang: Alexandr Wang is Chief AI Officer at Meta and founder of Scale AI, where he contributes to strategic AI initiatives and model development. He is recognized for his expertise in scaling AI systems and data infrastructure. His involvement in the Muse Spark 1.1 result highlights executive-level contributions to Meta’s coding agent advancements.
Muse Spark 1.1: Muse Spark 1.1 is an AI coding agent designed for high-performance task execution with emphasis on cost efficiency. It was evaluated in the Opencode harness on the Artificial Analysis Coding Agent Index, where it demonstrated competitive results relative to other frontier models. The project reflects collaborative efforts from Meta’s AI teams and associated leaders.
`json
{
“AI Research”: “Meta is focusing on achieving a strong blend of performance and cost-efficiency in its advanced AI coding agents.”,
“Leadership Engagement”: “Meta’s leadership is acknowledged for their involvement in promoting the performance of their AI models.”
}
`
