Leaderboard

FinSearchComp

Expert-written financial search-and-reasoning questions in three tiers: time-sensitive data fetching, simple historical lookup and complex historical investigation. There are Global and Greater China subsets.

Best reported

68.9

grok-4

Global subset average accuracy %

16 Sept 2025

FinSearchComp arXiv v1 (2025-09-16); HF dataset last modified 2025-10-27

Global subset average accuracy %, higher is better. 10 models.

Updated 16 Sept 2025

  1. grok-468.9
  2. GPT-5-Thinking63.9
  3. Gemini2.5-pro42.6
  4. DouBao39.1
  5. Gemini2.5-pro (Google Search)38.2
  6. Qwen3-235B-A22B-250737.4
  7. YuanBao-DeepSeek-V330.5
  8. DouBao-Thinking29.8
  9. HunYuan-T1-Thinking29.8
  10. YuanBao-DeepSeek-R129.8

FinSearchComp arXiv v1 (2025-09-16); HF dataset last modified 2025-10-27

Global subset average accuracy %, higher is better. 10 models.

Updated 16 Sept 2025

  1. 1grok-468.9
  2. 2GPT-5-Thinking63.9
  3. 3Gemini2.5-pro42.6
  4. 4DouBao39.1
  5. 5Gemini2.5-pro (Google Search)38.2
  6. 6Qwen3-235B-A22B-250737.4
  7. 7YuanBao-DeepSeek-V330.5
  8. 8DouBao-Thinking29.8
  9. 8HunYuan-T1-Thinking29.8
  10. 8YuanBao-DeepSeek-R129.8
FinSearchComp leaderboard (Global subset average accuracy %, higher is better)
1grok-4xAI68.9web productPaper Table 5
Reported by the authors, 16 Sept 2025
2GPT-5-ThinkingOpenAI63.9web productPaper Table 5
Reported by the authors, 16 Sept 2025
3Gemini2.5-proGoogle42.6web productPaper Table 5
Reported by the authors, 16 Sept 2025
4DouBaoByteDance39.1web productPaper Table 5
Reported by the authors, 16 Sept 2025
5Gemini2.5-pro (Google Search)Google38.2API with searchPaper Table 5
Reported by the authors, 16 Sept 2025
6Qwen3-235B-A22B-2507Qwen37.4web productPaper Table 5
Reported by the authors, 16 Sept 2025
7YuanBao-DeepSeek-V3Not stated30.5web productPaper Table 5
Reported by the authors, 16 Sept 2025
8DouBao-ThinkingByteDance29.8web productPaper Table 5
Reported by the authors, 16 Sept 2025
8HunYuan-T1-ThinkingTencent29.8web productPaper Table 5
Reported by the authors, 16 Sept 2025
8YuanBao-DeepSeek-R1Not stated29.8web productPaper Table 5
Reported by the authors, 16 Sept 2025

FinSearchComp arXiv v1 (2025-09-16); HF dataset last modified 2025-10-27

Greater China subset average accuracy %, higher is better. 10 models.

Updated 16 Sept 2025

  1. DouBao54.2
  2. YuanBao-DeepSeek-R152.5
  3. grok-451.9
  4. HunYuan-T1-Thinking50.5
  5. DouBao-Thinking49
  6. YuanBao-DeepSeek-V348.8
  7. GPT-5-Thinking46.4
  8. ERNIE-X140.8
  9. DeepSeek-R140.5
  10. Kimi k238.3

FinSearchComp arXiv v1 (2025-09-16); HF dataset last modified 2025-10-27

Greater China subset average accuracy %, higher is better. 10 models.

Updated 16 Sept 2025

  1. 1DouBao54.2
  2. 2YuanBao-DeepSeek-R152.5
  3. 3grok-451.9
  4. 4HunYuan-T1-Thinking50.5
  5. 5DouBao-Thinking49
  6. 6YuanBao-DeepSeek-V348.8
  7. 7GPT-5-Thinking46.4
  8. 8ERNIE-X140.8
  9. 9DeepSeek-R140.5
  10. 10Kimi k238.3
FinSearchComp leaderboard (Greater China subset average accuracy %, higher is better)
1DouBaoByteDance54.2web productPaper Table 5
Reported by the authors, 16 Sept 2025
2YuanBao-DeepSeek-R1Not stated52.5web productPaper Table 5
Reported by the authors, 16 Sept 2025
3grok-4xAI51.9web productPaper Table 5
Reported by the authors, 16 Sept 2025
4HunYuan-T1-ThinkingTencent50.5web productPaper Table 5
Reported by the authors, 16 Sept 2025
5DouBao-ThinkingByteDance49web productPaper Table 5
Reported by the authors, 16 Sept 2025
6YuanBao-DeepSeek-V3Not stated48.8web productPaper Table 5
Reported by the authors, 16 Sept 2025
7GPT-5-ThinkingOpenAI46.4web productPaper Table 5
Reported by the authors, 16 Sept 2025
8ERNIE-X1Baidu40.8web productPaper Table 5
Reported by the authors, 16 Sept 2025
9DeepSeek-R1DeepSeek40.5web productPaper Table 5
Reported by the authors, 16 Sept 2025
10Kimi k2Moonshot AI38.3web productPaper Table 5
Reported by the authors, 16 Sept 2025

Full details for FinSearchComp