Software engineering
APEX-SWE
200 integration and observability cases covering software construction and debugging with operational telemetry.
About APEX-SWE
What it measures
Task pass rate under the declared harness; expert rubrics and executable tests.
Scope & limitations
The publisher releases 50 related cases; its 200-case leaderboard split stays private. The public subset cannot establish a full-suite score.
Evaluation environment
Pinned software environments, task tools, unit tests and the publisher's evaluation harness.