Personalized benchmark noahbench re-released, crowning Fable 5.1, Grok 4.7 and Opus 5.5

sarahcat21 · x · 2026-09-29

Developer Noah re-released noahbench™, a deliberately personalized benchmark that grades model output quality on real-world tasks he solves daily.
His top 3 models: Fable 5.1, Grok 4.7, and Opus 5.5.
Quoting it, Sarah Cat argues the industry is entering an era of customization — per-customer and personalized models — which will require new modeling approaches plus tools to scale benchmarks/evaluation and to monitor, update, and orchestrate dozens to thousands of models.

More from Models