Unison - Internal Consistency: leaderboard
The internal consistency score of Unison (2026), a benchmark for unified multimodal models: whether what a model understands about a scene agrees with what it generates for it, as a unified percent score from the Unison-Judge and task metrics. Higher is better.
Metric: Unified score (%) on the internal consistency dimension of Unison (2,169 unified understanding-and-generation samples): whether what the model understands about a scene agrees with what it generates for it, scored by the authors' Unison-Judge (a human-aligned 8B judge) and task metrics; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 86.9 | #137 |
| 2 | GPT-5.2 | 84.7 | #176 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=unison-internal-consistency · How It Works · Data refreshed daily, snapshot 2026-11-13.