Explore the benchmark

Follow a question. Compare the options. Inspect the evidence.

Which configurations deserve a closer look?

Inspect every configuration

Quality, cost, and time describe separate parts of the result.

Sort table

Highest-score share counts tasks where this option scored best; tied wins are divided equally. Reused input is the share of input the AI could retrieve from cache. Neither is a task success rate.

Route metrics for the selected comparison
Quality rank	Route configuration	Quality score / 100	Highest-score share	Estimated AI cost / task	Reused input	AI time

Which results were included?

Task evidence

Inspect the underlying results.

See how each AI setup performed on individual tasks. Descriptions are anonymized to protect the source work.

Choose where to investigate

Browse the task records and available results from this run. These views change the order of tasks; your filters stay in place.

Back to comparison

Task listPublished task results

Engineering area Complexity Sort tasks

Published records

Tasks

Select a task to inspect its results

Anonymized task evidence for the selected official benchmark run.
Task	Task record	Area	Complexity	Highest-scoring setup	Scores across setups (0–100)	High–low gap

—