MTVA-Bench - Actions: 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 deterministic required/prohibited tool-action checks from 2609.20152v1, S7.T5: Score (0-100). Higher is better.
Metric: Score (0-100). Source: arxiv.org. Saturation forecast: Around November 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.6 Flash | 68.3 | #59 |
| 2 | GPT-5.2 | 65.5 | #116 |
| 3 | Qwen 3.8 27B | 64.2 | #125 |
| 4 | GPT-5.6 Sol | 63.2 | #22 |
| 5 | GPT-5.6 Luna | 58.6 | #80 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #6 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mtva-bench-actions · How It Works · Data refreshed daily, snapshot 2026-11-07.