Deep FinResearch Bench (Firm B Template): leaderboard
Deep FinResearch Bench (2026) tests AI agents on professional equity research: deep research agents and LLMs with web search write pre-earnings reports on S&P 500 companies that mirror brokerage analyst reports. Mean overall report quality from 1 to 4 judged by GPT-5, with the report rubric of the second brokerage (Firm B); higher is better.
Metric: Mean overall report-quality score (1-4 scale, 1 poor to 4 excellent) that GPT-5 assigns to AI-written pre-earnings equity research reports on S&P 500 companies (fiscal 2025 Q1 and Q2), each prompted with a rubric of Firm B's professional reports for the same company and quarter; averaged over comprehensiveness, assumption quality, coherence and analytical depth; deep research agents and LLMs with web search; higher is better. Source: arxiv.org. Saturation forecast: Around May 2028. 9 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 2.5 Pro Deep Research | 1.78 | ||
| 2 | Perplexity Deep Research | 1.68 | ||
| 3 | O3 Deep Research | 1.57 | ||
| 4 | Grok4 Deep Research | 1.55 | ||
| 5 | Sonar Pro | 1.54 | #743 | |
| 6 | Sonar Reasoning Pro | Thinking | 1.47 | #537 |
| 7 | GPT-4o | 1.4 | #754 | |
| 8 | GPT-4o Mini | 1.32 | #1680 | |
| 9 | Gemini 2.5 Flash | 1.25 | #466 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=deep-finresearch-bench-firm-b-template · How It Works · Data refreshed daily, snapshot 2026-11-22.