impulse

Watchlist

The models your team runs, on the hosts you run them on: when each one retires, what to move to, and what it costs. Nothing is stored on our side. The list lives in the link, and this browser remembers the last list you edited.

Start with the models you run

Search above, pick the host you call the model on and, if you pin versions, the exact version or region. Each one you add shows its retirement date on that host, what to move to and its price.

Try it with an example list

Gemini 3.6 Flash on Databricks · o3-pro on Azure OpenAI · Claude Sonnet 4.5 on Anthropic · GPT-6.2 Astra on Azure OpenAI · Claude Fable 5.2 on Amazon Bedrock · Gemini 4 Neon on Vertex AI

How it works

Retirement dates are the hosts’ own, as on Retiring, checked against the notices of OpenAI, Anthropic, Google, Microsoft and Amazon on 26 Oct 2026. A model you watch on a host follows every version and region the host lists; when only some of them retire, the row says which. Pin one version or region to follow just that one.

“Move to” is the newest model in the same line that no API retires within three months of the date, with whether your host serves it and how its price compares with what you pay now.

Prices are list prices per million tokens. Price moves come from the last 30 days of price history.