PAST-Bench - Information Gathering: leaderboard

Metric: Capability task score (%; information-gathering families; persistence on: the agent may reuse memory records, skills, profile state and session history produced by earlier episodes of the family; 26 scenarios and 204 episodes run as ordered sequences of fresh-session tasks; each evaluation episode scores completion (audit-data checks, or a MiniMax-M2.7 rubric judge for open-ended families), tool-error recovery and a safety gate; family-balanced means over three trials; Hermes agent framework). Source: arxiv.org. Saturation forecast: Around December 2027. 7 models tracked.

Top models

#ModelScore
1MiniMax-M2.771
2Kimi K2.6 (Thinking)70
3Claude Sonnet 4.669
4GLM-5.169
5GPT-5.4 (Medium)69
6Claude Opus 4.6 (Medium)68
7DeepSeek V4 Pro (High)66

Interactive version: theaggregate.ai/benchmark?slug=past-bench-information-gathering · How It Works · Data refreshed daily, snapshot 2026-09-29.