Feature · Admin

Benchmarks

Compare providers and inspect benchmark evidence.

Last changed

Generated from commit be57775 ·

How it works

Admin overview

Admin overview links every administrator tool: benchmarks and evaluation, people and access, models and routing, governance and spend, and, for platform operators, platform operations.

Benchmarks

Benchmarks has two tabs: Provider benchmarks compares model providers and CLIs on the same cases; Agent harness shows harness runs. Paid runs need consent from an owner or admin.

Who can open it

Only owners and admins. Others see "Admin access required".

From the help article The Admin area, checked against the product on .

What you can do

The operations this feature's screens call. Each one is POST /api/<name>, and the badge says who may call it.

  • harnessBenchmarks/listCases

    AdminNo API tokens

    Harness benchmarks: lists cases.

    Types: HarnessBenchmarkListCasesRequest returns HarnessBenchmarkListCasesResponse

  • harnessBenchmarks/listRuns

    AdminNo API tokens

    Harness benchmarks: lists runs.

    Types: HarnessBenchmarkListRunsRequest returns HarnessBenchmarkListRunsResponse

  • harnessBenchmarks/saveCase

    AdminNo API tokens

    Harness benchmarks: saves the case.

    Types: HarnessBenchmarkSaveCaseRequest returns HarnessBenchmarkCase

  • harnessBenchmarks/compareRuns

    AdminNo API tokens

    Harness benchmarks: compares runs.

    Types: HarnessBenchmarkCompareRunsRequest returns HarnessBenchmarkComparison

  • harnessBenchmarks/startRun

    AdminNo API tokens

    Harness benchmarks: starts the run.

    Types: HarnessBenchmarkStartRunRequest returns HarnessBenchmarkStartRunResponse

  • harnessBenchmarks/recordRun

    AdminNo API tokens

    Harness benchmarks: records a run.

    Types: HarnessBenchmarkRecordRunRequest returns HarnessBenchmarkRun

Screens

What it records

The feature writes these events to the audit log, the permanent record of what happened.

  • Harness benchmark case version created
  • Harness benchmark run recorded

Generated from code at commit be57775 ·

Turn the numbers into shipped work.

Agent runs these choices for you: a persistent AI worker with memory and rules, on your Claude and Codex subscriptions.