Think-Reason-Learn Ensemble on VCBench
Think-Reason-Learn Ensemble scores F0.5 37.9 on VCBench, the venture capital benchmark from the University of Oxford and Vela Research, with 40.6% precision and 30.1% recall on the 4,500-founder private test set. That is rank 1 of 32, 3.5× the F0.5 of tier-1 VCs and 4.4× Y Combinator, at $0.87 per 1,000 founders scored.
- Rank
- #1 of 32
- F0.5
- 37.9
- Precision
- 40.6%
- Recall
- 30.1%
- Cost / 1k founders
- $0.87
Scored on the 4,500-founder private test set, mean over three folds. That is 3.5× the F0.5 of tier-1 VCs. See the full leaderboard.
Compared with the reference rows
| Entry | Precision | Recall | F0.5 | Cost / 1k |
|---|---|---|---|---|
| Think-Reason-Learn Ensemble | 40.6% | 30.1% | 37.9 | $0.87 |
| Tier-1 VCs | 23.0% | 5.2% | 10.7 | n/a |
| Y Combinator | 14.0% | 6.9% | 8.6 | n/a |
| Random Classifier | 9.0% | 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
A mix of families: LLM-written rules and classical models are combined into one score, so an LLM still runs at scoring time. Think-Reason-Learn Ensemble used about 20,800 input and 7,448 output tokens per founder on Typesafe System One (jev-1.13.0), measured from the API usage fields. Cost is tokens times the provider's list price on 2026-09-24 ($0.042 input and $0 output per million tokens), so it recomputes when prices change. Build costs such as question or policy generation are excluded.
Read more
- VCBench Leaderboard Update: Accuracy vs Cost, Plus GPT-6 and Claude Opus 5.5: VCBench now plots F0.5 against the cost of scoring 1,000 founders, with GPT-6, Claude Opus 5.5, new submissions and the Think-Reason-Learn Ensemble.
Frequently asked questions
- What does Think-Reason-Learn Ensemble score on VCBench?
- Think-Reason-Learn Ensemble scores F0.5 37.9 on VCBench, with 40.6% precision and 30.1% recall, rank 1 of 32 as of 2026-10-07.
- Does Think-Reason-Learn Ensemble beat human investors at predicting founder success?
- Yes. Its F0.5 is 3.5 times that of tier-1 VCs (10.7) and 4.4 times Y Combinator (8.6), after both are normalized to the dataset's 9% base rate.
- How much does Think-Reason-Learn Ensemble cost per 1,000 founders?
- About $0.87 per 1,000 founders at list prices on 2026-09-24. A mix of families: LLM-written rules and classical models are combined into one score, so an LLM still runs at scoring time. Think-Reason-Learn Ensemble used about 20,800 input and 7,448 output tokens per founder on Typesafe System One (jev-1.13.0), measured from the API usage fields. Cost is tokens times the provider's list price on 2026-09-24 ($0.042 input and $0 output per million tokens), so it recomputes when prices change. Build costs such as question or policy generation are excluded.