Focused model compare

Jabberwock

Claude Fable 5 is the starting run, compared with DeepSeek V4 Flash, Kimi K3, GPT-5.6 Sol.