Focused model compare

Jabberwock

Claude Composer 2.5 is the starting run, compared with Kimi K3, GPT Sol 5.6 High, Grok 4.5.