Focused model compare

Jabberwock

Grok 4.5 is the starting run, compared with DeepSeek V4 Flash, Kimi K3, GPT-5.6 Sol.