Meta’s Muse Image climbs to second place in Arena’s image generation rankings, intensifying the AI arms race Meta's Muse Image model rose to second place on Arena.ai's Text-to-Image Arena leaderboard with an Elo score of 1280, trailing only OpenAI's GPT Image 2 at 1385. The model, announced by Meta Superintelligence Labs, features agentic tool use and emergent self-refinement, and is available via Meta AI app, Instagram Stories, and WhatsApp. This intensifies the AI arms race among major tech companies. Meta’s Muse Image climbs to second place in Arena’s image generation rankings, intensifying the AI arms race The new model from Meta Superintelligence Labs trails only OpenAI's GPT Image 2, signaling a serious escalation in the battle for AI-generated media dominance. Meta just crashed the AI image generation party, and it brought receipts. The company’s new Muse Image model has landed the number two spot on Arena.ai’s Text-to-Image Arena leaderboard, posting an Elo score of 1280 based on 7,715 community votes. Only OpenAI’s GPT Image 2, sitting comfortably at 1385, ranks higher. What Muse Image actually does Announced on July 7, 2026, by Meta Superintelligence Labs, the Muse Image model features agentic tool use, meaning it can actively leverage web search and coding functionalities during the image creation process. The model also exhibits what Meta calls emergent self-refinement during generation, using test-time compute scaling to improve results before you ever see the final image. Muse Image also secured the number two position in both the single-image edit and multi-image edit arenas on the same platform. Its companion model, Muse Video, grabbed the number three spot in Arena.ai’s Text-to-Video rankings. Where you can actually use it Meta has made Muse Image immediately accessible to consumers. The model is available through the Meta AI app, Instagram Stories in the United States, and WhatsApp in select countries. The competitive landscape just got more interesting An Elo gap of 105 points between Muse Image and GPT Image 2 is notable but not insurmountable, especially for a model in its initial release. For context, Elo systems are designed so that a difference of about 100 points translates to roughly a 64% expected win rate for the higher-rated player. The fact that the model emerged from Meta Superintelligence Labs suggests serious institutional commitment to competing directly with OpenAI and Google DeepMind at the highest level. Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .