Overview
Provider, model version, effort level, context window, native harnesses, tool support, cost estimate, and speed estimate.
Models
Each model page will show provider, version, effort level, context window, tool support, cost, speed, WYBench runs, known failure modes, verified evidence, and a clearly labeled Brandon personal-opinion note when available.
No model detail pages are published yet.
WYBench will not publish model claims, scores, strengths, weaknesses, or trust rankings until they are backed by verified runs or clearly cited public references.
WYBench will not publish model claims, scores, strengths, weaknesses, or trust rankings until they are backed by verified runs or clearly cited public references.
Model Page Structure
WYBench Results
Overall score, category scores, previous-version comparison, evidence links, and verification status.
Failure Modes
Assumptions, over-editing, instruction drift, fake completion, weak terminal recovery, broken tests, or harness dependence.