# IPO Finance Agent leaderboard

IPO due-diligence questions on recent S-1 filings, answered by a tool-using agent with web search and EDGAR access (forked from Vals AI's Finance Agent v2 harness).

## IPO Finance Agent: mean rubric score % (higher is better, 6 models)

| Rank | Model | Provider | Score | Metric | Setting | Reported by | Source | URL | Date |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| 1 | GLM-5.2 | Zhipu AI | 79.8 | mean rubric score % | public SpaceX set (n=70), agent harness | Reported by the authors | Paper Table 1 | https://arxiv.org/html/2606.23032v3 | 30 Jun 2026 (checked 1 Nov 2026) |
| 2 | GLM-5.1 | Zhipu AI | 79.7 | mean rubric score % | public SpaceX set (n=70), agent harness | Reported by the authors | Paper Table 1 | https://arxiv.org/html/2606.23032v3 | 30 Jun 2026 (checked 1 Nov 2026) |
| 3 | Qwen3.7-Max | Qwen | 78.9 | mean rubric score % | public SpaceX set (n=70), agent harness | Reported by the authors | Paper Table 1 | https://arxiv.org/html/2606.23032v3 | 30 Jun 2026 (checked 1 Nov 2026) |
| 4 | MiMo-2.5 Pro | Xiaomi | 77.2 | mean rubric score % | public SpaceX set (n=70), agent harness | Reported by the authors | Paper Table 1 | https://arxiv.org/html/2606.23032v3 | 30 Jun 2026 (checked 1 Nov 2026) |
| 5 | Kimi K2.6 | Moonshot AI | 68 | mean rubric score % | public SpaceX set (n=70), agent harness | Reported by the authors | Paper Table 1 | https://arxiv.org/html/2606.23032v3 | 30 Jun 2026 (checked 1 Nov 2026) |
| 6 | Nemotron 3 Ultra | NVIDIA | 45.5 | mean rubric score % | public SpaceX set (n=70), agent harness | Reported by the authors | Paper Table 1 | https://arxiv.org/html/2606.23032v3 | 30 Jun 2026 (checked 1 Nov 2026) |

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

## About this benchmark

- Maker: Mostapha Benhenda
- Year: 2026
- Use case: Financial filings Q&A
- Size: 1,000 questions
- Public split: 70 questions on the SpaceX (SPCX) S-1. The rest, on further S-1s across sectors, are private.
- Grading: Mean rubric score from 0 to 1 per question, with partial credit. Rubrics come from an automated generator and each is reviewed by a human expert.
- Licence: MIT (code, per GitHub repo)
- Submission: Not stated
- Products allowed: Not stated
- Latest version: arXiv v3 (2026-06-30)
- What it does not cover: Only S-1/IPO filings. Public results cover one company. No live market data or advice quality.
- Last checked: 1 Nov 2026
- Links: paper https://arxiv.org/abs/2606.23032, code https://github.com/benstaf/ipoagent
- Directory page: https://benchmarks.finance/benchmarks/ipo-finance-agent/
- Leaderboard: https://benchmarks.finance/leaderboards/ipo-finance-agent/

Source: https://benchmarks.finance/leaderboards/ipo-finance-agent/

Created and maintained by Moneda.