What it is
VCBench
VCBench is the first AGI benchmark for venture capital, built by Vela Partners: 9,000 anonymized historical founder profiles with a public leaderboard. It tests models on frozen history.
Every profile is anonymized, with 90%+ re-identification risk reduction, and state-of-the-art LLMs outperform Y Combinator and tier-1 firms on it.
How it works
In four parts
- I
The profiles
9,000 anonymized founder profiles, each with a ground-truth success label, and a fixed task definition every model is asked the same way.
- II
The score
F0.5 is the primary metric. Venture capital is a rare-event problem with asymmetric costs: missing a good founder is expensive, funding a bad one is much more so.
- III
The leaderboard
Public baselines for human VC firms and frontier LLMs, side by side, at vcbench.com.
- IV
What comes next
VC Arena moves to the live future: models predict startup-world events before they resolve, so nothing can be memorized, and calibration is measured alongside accuracy.
In numbers
- 9,000
- anonymized founder profiles
- 29.1%
- precision for vanilla GPT-4o, 3.2x the index
- 90%+
- re-identification risk reduction














Benchmarks
Simulation
Harness
Predictions