Meta's Muse Glimmer Debuts on Code Arena: WebDev Benchmark
Meta AI has released Muse Glimmer, a new 30B open-weight agentic model, which debuted at #77 overall and #26 among open models on the Code Arena: WebDev benchmark with 1,359 points. As an open-weight model designed to run locally on a single GPU, Muse Glimmer positions Meta among the top US labs offering open-weights models capable of handling agentic workflows. The model is Apache 2.0 licensed and specifically optimized for tool use, long-context tasks, and failure recovery. It was trained using a combination of supervised fine-tuning, on-policy distillation, and reinforcement learning.
## BACKGROUND
Code Arena: WebDev is a community-driven benchmark where users compare generated web applications from anonymous models to rank them using Elo ratings. Unlike SWE-bench, which focuses on bug fixing in existing codebases, WebDev Arena evaluates a model's ability to generate new front-end applications from scratch using React, TypeScript, and Tailwind CSS.