A side by side demo shows Gemini 4 Argon and Meta's Muse Wonder building very different games from the same prompt
Gemini 4 Argon vs Muse Wonder in game development. Same prompt, two completely different games.
Context
Informal head to head demos spread quickly after big launches and shape early impressions of new models, even without any controlled testing.
This is a single informal demo from the universalmodhq account, not a benchmark, and the prompt is not shown. Gemini 4 Argon launched on October 13, and Meta's Muse Wonder was covered on October 12.
Deep dive
On October 13, the universalmodhq account posted: Gemini 4 Argon vs Muse Wonder in game development. Same prompt, two completely different games.
The video compares two versions of what appears to be the same boss fight, in which a hooded knight battles a giant clockwork colossus called Horologe. The half labeled Gemini 4 Argon shows detailed 3D graphics with lighting, a clock tower, blossoming trees and styled health bars. The half labeled Muse Wonder shows the same scene with simple blocky shapes, flat colors and basic trees.
The post does not share the prompt, the tools or game engine used, how many attempts each model got, or whether any assets were added by hand, so it cannot show which model is better at game development in general.
Google launched Gemini 4 Argon broadly on October 13 and says it is built for complex coding and agentic problems. Meta introduced its Muse Wonder family of frontier models the day before. Other early comparisons in the decks, including Theo's bug fix test, have shown mixed results across models.
What to expect next
Watch whether controlled comparisons of game generation, with shared prompts and settings, back up the gap in this demo.
Also watch whether Meta or Google respond with their own game building showcases, and whether developers adopt either model for game prototyping in tools like Antigravity or Muse Code over the coming weeks. A shared prompt would let others repeat the test fairly.