Summary
Vercel added JSON-valued feature flags, letting teams store full configuration objects instead of single booleans, strings, or numbers. The changelog explicitly frames the feature as a cleaner way to A/B test AI models and swap model settings without scattering multiple flags across an app.
What changed
Vercel Flags gained support for JSON values, including model configuration objects with parameters like model ID, temperature, max tokens, and prompts.
Why it matters
This is a small platform primitive with real downstream impact for AI apps. Teams increasingly need to steer model choice and prompt configuration as a unit, and Vercel is making that operational pattern first-class inside its deployment platform.
Evidence excerpt
Vercel says JSON values let teams collapse several related AI flags into a single feature flag and use it to route traffic to different models.