What it measures
Grounded commercial judgement.
Synthetic cases, standard prompts, anchored rubrics, representative outputs and a leaderboard shaped by investment relevance.
About
A careful instrument for machine-era judgement.
MidMarketBench evaluates whether frontier models can produce useful investment workflow outputs from synthetic private-market cases.
What it does not measure
Not market truth.
No company is real. No score is an observed model result in this pass. The shell demonstrates the evaluation contract before live runs begin.
Provenance
Synthetic by construction.
Cases are built from realistic patterns, not real companies or proprietary deal materials. Model names and provider logos were checked against primary provider sources on 21 June 2026. Scores, ranks, release months and unspecified technical metadata are illustrative until run artefacts and provider snapshots are persisted.
Audience
Investment teams
PE and growth investors evaluating model usefulness in diligence and origination.
AI leads
Internal teams selecting models, prompts, tools and evaluation controls.
Model providers
Vendors testing whether general capability translates into decision-useful work.