Fable 5.1 on the NYC prompt.
Opus 5 underneath. It told me straight out this is not the best it can do, that going further would cost too many resources. First time a model has said that to me on this test. You can see it. 1,292 lines against Opus 5s 11,653. The camera never
28 Opus 5 subagents on max effort doing coding, research, review, 4M tokens so far, it still hasn't fixed the bug.
In the last days I've been trying different agentic configurations on Claude, I'm even building a multiagent orchestration plugin using Anthropic's stack, which is
Same brief. Same night. Same system prompt.
Left is Opus 5. Right is Fable 5.1. The brief was one line: build a dashboard for these 8 metrics. Four of the metrics are null. Opus drew a wave. Fable drew a sphere. Both left the four missing numbers missing. I ran every brief
Fable 5.1 on the NYC prompt.
Opus 5 underneath
You have reached the end of the archive
All of Claude Opus 5