Harness Rank
Task outcomes across model–harness combinations.
- 3 harnesses in scope
- 24 task specifications
- No measurements
- Draft 0.3.0
No trial records
No executed trials or result artifacts. Task specifications are available for review.
- Harness scope
- Codex · Hermes · Pi
- Runtime qualification
- Not run
- Task specifications
- 24 authored · workspace packages and graders pending
Draft · no measurements.
说明
Same model, provider route, tasks, and limits. Native harness instructions and context handling remain part of the comparison.
Proposed model: gpt-6-sol. Releases are not pinned; runtime compatibility is unqualified. Task specifications are authored; workspace packages and graders are pending.