Model speed, in context.

Explore response throughput and elapsed time across saved benchmark runs. Select a measurement to see the runs behind it.

A snapshot of completed runs, not a live benchmark. Results may change with the task and its version.

Models

Chart effort

Charts follow these controls. The table shows all efforts.

Response speed

tokens / second

Mean response throughput by model and speed setting.

StandardFast

Speed and wall time

same runs

Mean elapsed seconds against mean response throughput.

StandardFast

Wall time · seconds

Model × effort

response tokens / second

Each value is the arithmetic mean of up to three latest completed measurable runs for the same task and verified task version. Counts show the number included. A dash means no measurable result. Select any cell for saved runs and failures.

Run details

Select a model and effort in the table or a chart to inspect its saved runs.