Haiku + Gemini + GPT-4o mini, persona

Name: SynthBench Leaderboard
Creator: SynthBench
License: https://opensource.org/licenses/MIT

ensemble

Ensemble of Haiku 4.5, Gemini 2.5 Flash-Lite, GPT-4o mini (all persona)

OpinionsQA · US adults

Index on OpinionsQA 48 above random answers (0); real people score 100

opinionsqa_ensemble_3blend_20260412_020745

Configuration

Ensemble · default settings

Method: Ensemble
Base provider: —
Model: 3-model-blend
Dataset: OpinionsQA
Temperature: default
Reasoning effort: default
Template: current
Samples/question: —
Timestamp: 2026-04-12T02:07:45Z

View all replicates for this config → Haiku + Gemini + GPT-4o mini, persona profile →

Aggregate scores

SynthBench Index: 0 = random answers, 100 = real people. SPS is the raw 0 to 1 score it is computed from. Index = 100 × (SPS − SPS of random answers) ÷ (human ceiling − SPS of random answers), on this run's dataset. The leaderboard averages it over the core datasets.

Loading scores…

Per-question detail All OpinionsQA questions, ranked by divergence →

Modal answer differs from human Topic

Per-question results for this run. Select a column header to sort; expand a row via the button in the Question column for full distribution bars.

Loading per-question detail…