EVA-Bench - Conversation Progression: leaderboard
Voice agents handle simulated phone calls from an audio user simulator in 213 enterprise scenarios (airline customer service, healthcare HR and IT service management), calling tools against a scenario database; EVA-Bench (Bogavelli et al., 2026). Conversation progression in percent: an LLM judge's 1-3 rating, rescaled to 0-100, of whether the agent moves the call toward resolution without repeating itself, losing context, stalling or backtracking, over five clean-audio trials per scenario, averaged equally over the three domains; higher is better.
Metric: Conversation progression (%; mean LLM-judge rating mapped from 1-3 to 0-1); five clean-audio trials, 213 scenarios, equal mean over three domains. Source: servicenow.github.io. Saturation forecast: Rough model projection: around 2026. 25 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Cartesia Managed Agents - GPT 5.5 | 82.06 | |
| 2 | Scribe v2.2 Realtime + Gemini 3 Flash + TTS Conversational v3 (ElevenAgents) | 80.42 | |
| 3 | Scribe v2.2 Realtime + GPT-5.4 + Eleven Flash v2 (ElevenAgents) | 76.61 | |
| 4 | Flux General Multi + GPT-5.4 + Sonic 3.5 | 76.03 | |
| 5 | Scribe v2.2 Realtime + Claude Haiku 4.5 + Eleven Flash v2 (ElevenAgents) | 75.08 | |
| 6 | Gemini 3.8 Live | 73.77 | |
| 7 | Universal 3.5 Pro + GPT-5.4 + Sonic 3.5 | 73.53 | |
| 8 | Ink 2 + Haiku 4.5 + Sonic 3.5 | 71.87 | |
| 9 | GPT Realtime 1.5 | 67.93 | |
| 10 | Apriel Voice 27B + Sonic 3.5 | 67.76 |
Interactive version: theaggregate.ai/benchmark?slug=eva-bench-conversation-progression · How It Works · Data refreshed daily, snapshot 2026-11-12.