Pseudonymous benchmark LisanBench ranks OpenAI GPT-5.6-Sol at 162 ECI, narrowly beating Anthropic Claude Fable 5 · Digg