Leni is a purpose-built AI analyst for commercial real estate and private equity — not a general chatbot pointed at finance, but an agentic system that handles underwriting, portfolio reporting, market research, and memo generation off data pulled from property and finance systems.
## The benchmark claim
What got it attention is accuracy. Leni placed first on the DRACO deep-research benchmark at 71.6%, ahead of the deep-research products from Perplexity, Google, and OpenAI. It finished top-two globally on SpreadsheetBench Verified, getting 365 of 400 tasks right. And on BullshitBench — which measures whether a model catches fabricated premises — it flagged 98% of them, ahead of all 142 public models on the leaderboard.
## Why vertical matters
General models hallucinate confidently on spreadsheets and invented premises, which is exactly where finance work breaks. Leni’s bet is that a system tuned for one domain and wired into the actual data sources beats a frontier generalist on the tasks that domain runs every day — underwriting a deal, not writing a poem about one.

Leave a comment