Focused model compare

Helm's Deep

Grok 4.6 is the starting run, compared with Claude Opus 5.5, GPT-6 Astra, GPT-6 Sol.