Model Retirement Radar is live
Steven · 2026-08-26
AI model endpoints change, become deprecated, and eventually retire. The new Model Retirement Radar puts those lifecycle changes next to a practical migration estimate.
FEATBrowse known lifecycle changes. The Radar separates retired and deprecated models, shows the vendor for every entry, and keeps the complete market list in one place.
Click any model in Known lifecycle changes to select it. The page moves directly to its Migration brief, so there is no separate search or setup step.
FEATBuild a migration brief. Enter the input tokens, requested output tokens, and expected monthly request count for your workload. The default scenario uses 20,000 input tokens, 2,000 output tokens, and 10,000 monthly requests. Inputs support localized number formatting and workloads up to one million input tokens.
The candidate table filters active models by their published context and maximum-output capacity. Each result includes a projected monthly estimate, the cost difference from the selected model, and a vendor-continuity marker when the replacement comes from the same provider.
Capacity and price are the starting point for a migration review. Test response quality, latency, tool support, regional availability, and compliance requirements before moving production traffic.
A quick workflow is: choose a deprecated or retired model, adjust the workload inputs, compare the lowest-cost compatible candidates, then shortlist same-vendor and cross-vendor options for evaluation.