Muse Spark 1.2 Now Available on Arena's Battle and Agent Modes
Arena has announced that Muse Spark 1.2, the latest version of Meta's multimodal reasoning model, is now available for public testing. Users can evaluate the model's performance in both side-by-side Chatbot Arena battles and autonomous agent workflows. This release allows the AI community to benchmark Meta's new model against other frontier models in real-world scenarios. Testing in Agent Mode is particularly significant as it measures the model's ability to act as an autonomous agent for complex, multi-step tasks. Muse Spark 1.2 can be tested on arena.ai, where its performance in Agent Mode will contribute to the Agent Arena Leaderboard. The model is designed to handle tasks such as browsing, coding, and researching by executing multi-step workflows from a single prompt.
## BACKGROUND
Muse Spark is a proprietary large language model developed by Meta Superintelligence Labs (MSL), designed for multimodal reasoning, tool-use, and visual chain of thought. Arena (formerly Chatbot Arena) is a popular benchmarking platform that ranks AI models based on human preference. Its Agent Mode specifically evaluates how well models can autonomously browse, research, and code to complete real-world tasks.