Model compare

City Scroll Journey

Compare the same prompt across GPT-6 Astra.

Missing runs for deepseek-v4-flash, KimiK3, gpt-sol-5-6-high on this benchmark.