SuperResearch v1 - Deep Research System Consistency: leaderboard
SuperResearch paper 2603.00582 v1, 300 tasks and 12 systems. This pinned experiment defines Overall as the aggregation of Coverage, Consistency, Utility and Objectivity. Current website metrics use a different definition and are not mixed with these results. Coverage is depth-weighted fact/insight/conclusion recall against expert knowledge anchors. Architecture: Deep Research System. Search-augmented baselines use the common LangGraph/Tavily scaffold. Native search and deep-research systems retain distinct endpoint identities; unspecified snapshots remain source-qualified. Consistency. Metric: Score (%).
Metric: Score (%). Source: arxiv.org. Saturation forecast: Estimated already saturated. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O4 Mini Deep Research | 20.04 | #229 |
| 2 | O3 Deep Research | 19.9 | #129 |
| 3 | Sonar Deep Research | 16.89 |
No result here: #5 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=superresearch-v1-deep-research-system-consistency · How It Works · Data refreshed daily, snapshot 2026-10-26.