MTVA-Bench - Criteria: leaderboard
MTVA-Bench evaluates the language-model component of cascaded voice agents on 490 authored scenarios for 49 agents in seven languages. Each scenario uses one rollout; caller-invalid episodes are excluded. Language subsets contain different agents and tasks, rather than matched translations. This series reports judged task criteria from 2609.20152v1, S7.T5: Score (0-100). Higher is better.
Metric: Score (0-100). Source: arxiv.org. Saturation forecast: Around February 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol | 82.7 | #29 |
| 2 | GPT-5.6 Luna | 66.2 | #118 |
| 3 | Gemini 3.6 Flash | 64.8 | #74 |
| 4 | Qwen 3.8 27B | 62.2 | #206 |
| 5 | GPT-5.2 | 59 | #174 |
No result here: #4 Claude Opus 5.5, #5 GPT-6 Astra, #9 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mtva-bench-criteria · How It Works · Data refreshed daily, snapshot 2026-11-08.