Overview

Available results, comparison conditions and evidence across all three boards.

Peer benchmark boards and their test coverage
Board	Comparison and recorded results	Conditions and evidence
AI Intelligence	Model × reasoning effort Core 1.0.7 · 17 historical configurations Sol · medium77.588Conditional 95% interval: 65.0 to 86.6. Sol · low76.358Conditional 95% interval: 63.5 to 85.7. Luna · max76.314Conditional 95% interval: 63.4 to 85.7. Ability score · top point estimates; lines show conditional 95% intervals. Current Core 1.1.0 has no Official results. Reading published results…	Same task set and scoring within each benchmark version. Open leaderboard Data Method Records
Harness Rank	Model × harness Design only · no measured attempts 24 task specifications · 12 planned workspaces Bug repair6 specifications Feature work6 specifications Local workflow6 specifications Refactor and migration6 specifications Inspect task specifications →	Plan: matched model, tasks and budget for Codex, Hermes and Pi. Open leaderboard Data Method Records
Mem	Memory systems, readers and writers 3 studies · 6 comparisons OpenViking11/13 RAGFlow rc111/13 Native systems · complete answers with source-matched quotes. Reader and writer tests remain separate.	Compare profiles within each retained test; workloads keep separate measures. Open leaderboard Data Method Records

Metrics belong to the selected benchmark. Metric definitions and evidence scope