AgentIdeaBench - Active: leaderboard
Research-hypothesis ideation in AgentIdeaBench's Active track, where the model gets a subfield name and a literature search tool and decides what to look up (arXiv 2609.07611v1). Weighted LLM-critic total on a 1-10 scale (originality 2, impact 1.5, feasibility 1, clarity 0.5, specificity 0.5; highest of three critics dropped), from reports/leaderboard_full.json; higher is better.
Metric: Weighted Hypothesis Score (1-10). Source: moyunxiang.com. Saturation forecast: Around June 2027. 63 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 7.14 | #16 |
| 2 | GPT-5.6 Sol | 6.88 | #22 |
| 3 | GPT-5.6 Terra | 6.8 | #40 |
| 4 | Claude Opus 4.8 | 6.79 | #39 |
| 5 | Claude Opus 4.7 | 6.7 | #51 |
| 6 | GPT-5.5 | 6.68 | #33 |
| 7 | Claude Sonnet 5 | 6.66 | #65 |
| 8 | GPT-5.2 | 6.66 | #115 |
| 9 | GPT-5.4 | 6.62 | #83 |
| 10 | GPT-5.6 Luna | 6.43 | #79 |
| 11 | GPT-5 | 6.4 | #105 |
| 12 | GPT-5 Mini | 6.38 | #193 |
| 13 | GPT-5.4 Mini | 6.36 | #228 |
| 14 | GLM-5.1 | 6.33 | #109 |
| 15 | Gemini 3 Flash (Preview) | 6.29 | #90 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=agentideabench-active · How It Works · Data refreshed daily, snapshot 2026-11-05.