Focused model compare

Helm's Deep

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