Time to REFLECT: leaderboard
Deep research agents increasingly automate complex information-seeking tasks, producing evidence-grounded reports via multi-step reasoning, tool use, and synthesis. Their growing role demands scalable, reliable evaluation, positioning LLM-as-judge as a supervision paradigm for assessing factual accuracy, evidence use, and reasoning quality. Yet the reliability of these judges for deep research age
Metric: Overall (Report Quality) (self-reported). Source: benchmarklist.com. Saturation forecast: Around August 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.3 Codex | 47.5 | #69 |
| 2 | GPT-OSS-120B | 46.5 | #341 |
| 3 | GPT-5 Mini | 43.5 | #186 |
| 4 | Qwen 3 32B | 39.5 | #443 |
| 5 | GPT-5.4 | 38 | #79 |
| 6 | Claude Opus 4.7 | 35 | #49 |
| 7 | Gemini 3.1 Pro (Preview) | 32.5 | #58 |
| 8 | Claude Haiku 4.5 | 32.5 | #280 |
| 9 | Qwen 3 235B A22B | 27 | #313 |
| 10 | Gemini 2.5 Flash | 24.5 | #245 |
| 11 | Gemma 3 27B | 20 | #607 |
| 12 | Qwen 3 8B | 14.5 | #674 |
| 13 | Llama 3.1 70B | 9.5 | #592 |
| 14 | Gemini 2.0 Flash | 4.5 | #342 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=time-to-reflect · How It Works · Data refreshed daily, snapshot 2026-10-20.