Each cell is sourced or an honest em-dash — never fabricated. Capability the tool’s own benchmark (SWE-bench Verified target; model = subcomponent via our Model Tracker) ·
Governance vendor /trust + /security pages (verify-or-em-dash; open-source ⇒ open-local) ·
Cost seat + compute (TCO) ·
Velocity npm / PyPI / VS Code / GitHub (free, durable).
We rank on the benchmark, never on our own tool. Confidence reflects data completeness until per-source ratings land; Governance / SWE-bench / TCO-compute fill in as their crons ship.