# FinDeepResearch leaderboard

Whether deep-research agents and LLMs can write a rigorous corporate financial analysis report that follows an expert-designed structure. Grading covers each analyst step: recognising data, calculating metrics, abstracting, and interpreting.

## FinDeepResearch: overall accuracy score (0-100) (higher is better, 10 models)

| Rank | Model | Provider | Score | Metric | Setting | Reported by | Source | URL | Date |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | FinAgentTest | Not stated | 41.9 | overall accuracy score (0-100) | Deep Research | Official board | OpenFinArena leaderboard | https://openfinarena.com/fin-deep-research | 23 Mar 2026 (checked 1 Nov 2026) |
| 2 | Gemini 3 Pro Deep Research | Google | 38 | overall accuracy score (0-100) | Deep Research | Official board | OpenFinArena leaderboard | https://openfinarena.com/fin-deep-research | 30 Dec 2025 (checked 1 Nov 2026) |
| 3 | OpenAI o3-deep-research | OpenAI | 37.9 | overall accuracy score (0-100) | Deep Research | Official board | OpenFinArena leaderboard | https://openfinarena.com/fin-deep-research | 15 Oct 2025 (checked 1 Nov 2026) |
| 4 | Grok 4 | xAI | 37.3 | overall accuracy score (0-100) | Deep Research | Official board | OpenFinArena leaderboard | https://openfinarena.com/fin-deep-research | 15 Oct 2025 (checked 1 Nov 2026) |
| 5 | Gemini 2.5 Pro | Google | 36.2 | overall accuracy score (0-100) | Deep Research | Official board | OpenFinArena leaderboard | https://openfinarena.com/fin-deep-research | 15 Oct 2025 (checked 1 Nov 2026) |
| 6 | OpenAI GPT-5 | OpenAI | 33 | overall accuracy score (0-100) | Thinking + Search | Official board | OpenFinArena leaderboard | https://openfinarena.com/fin-deep-research | 15 Oct 2025 (checked 1 Nov 2026) |
| 7 | Gemini 3 Pro | Google | 32.7 | overall accuracy score (0-100) | Thinking + Search | Official board | OpenFinArena leaderboard | https://openfinarena.com/fin-deep-research | 30 Dec 2025 (checked 1 Nov 2026) |
| 8 | Tongyi Deep Research | Qwen | 32.1 | overall accuracy score (0-100) | Deep Research | Official board | OpenFinArena leaderboard | https://openfinarena.com/fin-deep-research | 15 Oct 2025 (checked 1 Nov 2026) |
| 9 | Claude Sonnet 4.5 | Anthropic | 25.8 | overall accuracy score (0-100) | Thinking + Search | Official board | OpenFinArena leaderboard | https://openfinarena.com/fin-deep-research | 15 Oct 2025 (checked 1 Nov 2026) |
| 10 | Perplexity Sonar Deep Research | Perplexity | 24.7 | overall accuracy score (0-100) | Deep Research | Official board | OpenFinArena leaderboard | https://openfinarena.com/fin-deep-research | 15 Oct 2025 (checked 1 Nov 2026) |

Scores are never combined across benchmarks. Each row names its metric, source and date.

## About this benchmark

- Maker: National University of Singapore, 6Estates, Asian Institute of Digital Finance, Hefei University of Technology, USTC (OpenFinArena)
- Year: 2025
- Use case: Analyst Financial research
- Size: 64 listed companies, 8 markets, 4 languages, 15,808 grading items (247 per report)
- Public split: Full set public on Hugging Face, including analytical-structure templates and expert-annotated reference reports
- Grading: HisRubric hierarchical rubric judged by an LLM. Recognition/calculation items 1 mark, abstraction/interpretation items 2 marks; accuracy score out of 350, reported 0-100. Separate structure score
- Licence: Apache-2.0 (data, HF); paper CC BY 4.0
- Submission: open
- Products allowed: Yes
- Latest version: Paper v2 (2026-01-07); latest board entry 2026-03-23
- What it does not cover: Report writing only; no spreadsheet or tool actions, trading or advice suitability
- Last checked: 1 Nov 2026
- Links: paper https://arxiv.org/abs/2510.13936, data https://huggingface.co/datasets/OpenFinArena/FinDeepResearch, code https://github.com/OpenFinArena/OpenFinArena, leaderboard https://openfinarena.com/fin-deep-research
- Directory page: https://benchmarks.finance/benchmarks/findeepresearch/
- Leaderboard: https://benchmarks.finance/leaderboards/findeepresearch/

Source: https://benchmarks.finance/leaderboards/findeepresearch/

Created and maintained by Moneda.