V
VCBench
LeaderboardHow it worksModelsPostsResearch using VCBench
Repositories
Think-Reason-Learn
Papers
VCBenchGPTreePolicy InductionRandom Rule ForestReasoned Rule Mining
V
VCBench
← All models

FinGPT-VC1 on VCBench

ColumbiaReasoning
Board as of 2026-10-07

FinGPT-VC1 scores F0.5 22.2 on VCBench, the venture capital benchmark from the University of Oxford and Vela Research, with 21.8% precision and 24.2% recall on the 4,500-founder private test set. That is rank 21 of 32, 2.1× the F0.5 of tier-1 VCs and 2.6× Y Combinator.

FinGPT-VC1 on the VCBench leaderboard today
Rank
#21 of 32
F0.5
22.2
Precision
21.8%
Recall
24.2%
Cost / 1k founders
n/a

Scored on the 4,500-founder private test set, mean over three folds. That is 2.1× the F0.5 of tier-1 VCs. See the full leaderboard.

Compared with the reference rows

EntryPrecisionRecallF0.5Cost / 1k
FinGPT-VC121.8%24.2%22.2n/a
Think-Reason-Learn Ensemble40.6%30.1%37.9$0.87
Tier-1 VCs23.0%5.2%10.7n/a
Y Combinator14.0%6.9%8.6n/a
Random Classifier9.0%9.0%9.0$0

Human rows are normalized to the dataset's 9% success rate, so every entry is compared on the same base rate. Costs use list prices on 2026-09-24.

How it was scored

An LLM reads each founder profile at scoring time and returns a prediction. The cost per 1,000 founders is not reported because the submitter did not share token counts, so this entry appears on the rankings but not on the cost chart.

Read more

  • FinGPT-VC: Financial Large Language Models for Founder Success Prediction in Venture Capital (Jingyu Huang, Sitong Zhu, James Tang, Xiao-Yang Liu, Columbia University). Fine-tunes financial LLMs on VCBench; FinGPT-VC1 and FinGPT-VC2 are on the leaderboard.

Frequently asked questions

What does FinGPT-VC1 score on VCBench?
FinGPT-VC1 scores F0.5 22.2 on VCBench, with 21.8% precision and 24.2% recall, rank 21 of 32 as of 2026-10-07.
Does FinGPT-VC1 beat human investors at predicting founder success?
Yes. Its F0.5 is 2.1 times that of tier-1 VCs (10.7) and 2.6 times Y Combinator (8.6), after both are normalized to the dataset's 9% base rate.
How much does FinGPT-VC1 cost per 1,000 founders?
Not reported. The submitter did not share token counts, so FinGPT-VC1 appears on the rankings but not on the cost chart.

About VCBench

VCBench is the first benchmark for venture capital. It tests how well AI models, AI-native venture capital methods and human investors predict which startup founders will succeed, on 9,000 anonymized founder profiles. It was built by the University of Oxford and Vela Research, the research arm of Vela Partners, an AI-native quant venture capital firm in San Francisco. Many methods on the leaderboard are open source in Think-Reason-Learn.