Retired benchmark. These historical results are preserved for inspection and excluded from active-suite run counts. Browse the archive →

Focused model compare

Low-Poly World Explorer

GPT-5.5 is the starting run, compared with GPT-6 Astra.