Product
Background catalogue refresh and faster model search
Model search now uses the local catalogue. Public API checks completed in about 65 to 159 milliseconds after the latest update.
Speeds up model catalogue search using local caching and cancellation of outdated requests.
The earlier background refresh update did not remove every delay in model search. The latest update also removes repeated catalogue processing and stops search from waiting for an unrelated provider refresh.
After deployment, public API checks for Qwen and DeepSeek in the full model selection completed in about 65 to 159 milliseconds. These checks measure the response, not the time until results appear on every device. Typing adds a short delay before a query is sent, and network conditions can affect loading.
Outdated search requests are cancelled so an older response cannot replace the current results. Model permissions, prices and the existing customer pause remain in effect.