Add latency and TPS counters to subscription models
OpenUpdated Sep 6, 20261 reply
Some of the subscription models take forever to start replying. Would be nice to know how fast they are in advance.
NanoGPT teamSep 6, 2026
Model details now show average generation speed (TPS) and time to first token (TTFT) where we have performance data. You can use these to compare subscription models before sending a message. Provider lists also show speed and latency for routes with available data. Actual performance can vary with load and your request.