The social media giant's new video generation model debuted with strong community ratings, landing just behind Google and ByteDance on arena.ai's blind-tested leaderboard.
Meta just crashed the text-to-video party, and it showed up dressed to compete. The company’s Muse Video model landed at rank three on arena.ai’s Text-to-Video Arena leaderboard with an Elo-style score of 1459 ±15, based on 2,152 community votes. For a model that was only announced on July 7, that’s a remarkably strong debut in a field where Google and ByteDance have been setting the pace.
The ranking puts Muse Video behind Google’s gemini-omni-flash, which holds the top spot at 1527, and ByteDance’s dreamina-seedance-2.0-720p at 1482. It ranks ahead of Alibaba’s happyhorse-1.0, meaning three of the top four text-to-video models now belong to trillion-dollar tech companies.
How the rankings actually work #
Arena.ai’s leaderboard runs on blind community votes. Users are shown outputs from different models without knowing which model produced what, then pick which result they prefer. The platform has accumulated over 479,075 total votes across models in its Text-to-Video section. That sample size gives the Elo-style rankings meaningful statistical weight.
What makes Muse Video different #
The standout feature is native audio support. Most text-to-video models generate silent clips, leaving creators to layer in sound separately. Muse Video generates audio alongside the visual output. Meta unveiled Muse Video alongside Muse Image on July 7, positioning both as products of its Superintelligence Labs division. The company has indicated plans to roll out Muse Video to creators and integrate it into Meta AI in the coming months.
Why this matters for markets and investors #
Every dollar Meta, Google, and ByteDance pour into generative AI infrastructure reinforces the centralized cloud computing paradigm. That creates an interesting tension with the decentralized AI narrative that projects like Render, Akash, and Bittensor are trying to build. There’s also the GPU demand angle: training and running models like Muse Video requires enormous computational resources, and sustained demand from the world’s largest tech companies has downstream effects on crypto mining economics, as GPU availability and pricing directly impact proof-of-work mining operations and the emerging AI-crypto compute marketplace.
Google holds the top position with a 68-point Elo lead over Muse Video. ByteDance sits in between. The broader trend is unmistakable: multimodal AI—models that handle text, image, video, and audio natively—is where the industry is heading. Meta arriving at rank three on its first attempt suggests Superintelligence Labs is producing real results.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our