PAST-Bench - Information Gathering: leaderboard
Part of PAST-Bench (Xue et al., 2026): task score in percent on the information-gathering task families, with persistence of earlier episodes switched on, in the Hermes agent framework. Scores combine task completion, tool-error recovery and a safety gate; higher is better.
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: Not forecast. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | MiniMax-M2.7 | 71 | #518 |
| 2 | Kimi K2.6 (Thinking) | 70 | #178 (Kimi K2.6) |
| 3 | Claude Sonnet 4.6 | 69 | #161 |
| 4 | GLM-5.1 | 69 | #179 |
| 5 | GPT-5.4 (Medium) | 69 | #123 (GPT-5.4) |
| 6 | Claude Opus 4.6 (Medium) | 68 | #102 (Claude Opus 4.6) |
| 7 | DeepSeek V4 Pro (High) | 66 | #188 (DeepSeek V4 Pro) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=past-bench-information-gathering · How It Works · Data refreshed daily, snapshot 2026-11-13.