Skip to main content

Live model performance metrics accessible via AI Gateway

By Steven Van ·

AI Gateway now shows hourly-updated P50 and P95 latency and throughput per model and provider, in the model list, detail pages, and a REST API.

AI Gateway, part of Vercel, now shows live throughput and latency data for hundreds of models, pulled from actual AI Gateway customer requests and refreshed every hour.

  • The model list has sortable latency and throughput columns, showing each model's best P50 time-to-first-token and token throughput across all providers that carry it.
  • Model detail pages break performance down by provider, so you can compare, for example, which provider serving openai/gpt-oss-120b is fastest. These only show providers with enough traffic to measure.
  • A REST API endpoint (ai-gateway.vercel.sh/v1/models/[creator/model-name]/endpoints) returns rolling P50 and P95 latency and throughput per provider for programmatic use.
Vercel
Vercel
The platform for frontend developers — deploy, preview, and scale web apps and AI agents with zero config.
View Vercel →

Read the original announcement →

Read Live model performance metrics accessible via AI Gateway on Creators Toolbox