Four frontier models given the same flight-simulator prompt, with Astra first and Gemini cheapest

OpenAI GPT-6 Astra, Moonshot AI Kimi K3, Anthropic Claude Fable 5.1, and Google Gemini 3.8 Flash were each given the same prompt to build a flight simulator, producing playable results that differed more in smoothness and cost than in basic completeness. An informal side-by-side test ranked Astra first overall, treated Kimi K3 and Fable 5.1 as roughly tied, and placed Gemini last on quality while noting that Gemini finished in about two to three minutes at a cost of a few cents.