MP-Bench - Understanding Accuracy: leaderboard
MP-Bench (2026) evaluates conversational voice agents as active participants in synthesized four-participant multiparty conversations. This board is understanding overall accuracy, comprehension question answering over the Discussion scenario judged by a GPT-5 LLM judge, in percent, higher is better.
Metric: Understanding overall accuracy (%; comprehension question answering over multiparty Discussion dialogues, GPT-5 judge). Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 73.3 | #91 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=mp-bench-understanding-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-12.