Voice AI agents ranked for live conversation
Compare speech-to-speech systems on conversational preference, then audit interruptions, corrections and confirmations.

👁 111 reads
A caller changes Friday to Thursday while the booking agent is speaking. A smooth voice can still confirm the wrong day. For a live US appointment line I would listen to the overlap and read the resulting appointment record before choosing the voice.
What the board measures
Cartesia Managed Agents leads this snapshot. The rows include managed systems and particular reasoning configurations. The vote measures the experience of a conversation. It does not give a phone number, telephony route, price for transferring to a person or guarantee of correct tool execution.
The November 17, 2026 snapshot
The table covers the top 25 and additional recently added entries captured in the source. Prices are the evaluator’s reported USD benchmark estimates. A blank price means the evaluator did not supply one.
| Rank and model | Elo and votes | Benchmark price |
|---|---|---|
| 1. Cartesia Managed Agents | 1111 (685 votes) | |
| 2. Gemini 3.1 Flash Live Minimal | 1089 (990 votes) | |
| 3. Gemini 3.8 Live | 1074 (606 votes) | |
| 4. Gemini 3.1 Flash Live High | 1062 (957 votes) | |
| 5. Grok Voice 3.0 | 1059 (668 votes) | |
| 6. GPT-Live-1 (Sol, low) | 1037 (773 votes) | |
| 7. GPT-Live-1 (Astra, medium) | 1027 (736 votes) | |
| 8. Phonic 1 (High) | 1012 (683 votes) | |
| 9. GPT-Realtime-1.5 | 1000 (921 votes) | |
| 10. Grok Voice 2.0 | 991 (669 votes) | |
| 11. Eleven Realtime v1 Max | 990 (676 votes) | |
| 12. ElevenLabs Agents (Default Cascaded System) | 985 (893 votes) | |
| 13. Eleven Realtime v1 Mini | 981 (661 votes) | |
| 14. GPT Realtime (Aug ’25) | 967 (917 votes) | |
| 15. Gemini 3.8 Live Extended Thinking (High) | 966 (773 votes) | |
| 16. Nova 2.0 Sonic (Mar 2026) | 964 (913 votes) | |
| 17. GPT-Realtime-2.1 Minimal | 933 (932 votes) | |
| 18. GPT-Realtime-2.1 High | 931 (887 votes) | |
| 19. GPT Realtime Mini (Oct ’25) | 922 (962 votes) | |
| 20. GPT-Realtime-2 (High) | 920 (905 votes) | |
| 21. GPT-Realtime-2 (Minimal) | 918 (980 votes) | |
| 22. Nova Sonic | 907 (827 votes) | |
| 23. Grok Voice 1.0 | 903 (864 votes) | |
| 24. Deepgram Voice Agent (Default Cascaded System) | 901 (955 votes) | |
| 25. Inworld Realtime (Default Cascaded System) | 884 (941 votes) |
Reading the result
Preference and cost. The score comes from the listening or viewing comparison. The price column describes a separate estimate. For a working budget I would record actual spend across all attempts and divide by the number of accepted deliverables.
For the next step, see the Quantslant voice agent workflow.
Changes since the October archive
The October 13 crawl contains 26 entries on this track. The November 17 board contains 30. The following labels occur only in the later capture. This comparison does not establish the exact release day or rule out a renamed entry.
- Cartesia Managed Agents
- Eleven Realtime v1 Max
- Eleven Realtime v1 Mini
- Grok Voice 1.0
- Grok Voice 2.0
- Grok Voice 3.0
- Phonic 1 (High)
Sources
- https://artificialanalysis.ai/speech-to-speech/arena (accessed November 17, 2026)
- https://artificialanalysis.ai/speech-to-speech/arena?utm_source=alphasignal (accessed November 17, 2026)
The link to ElevenLabs is a partner link: if you buy through it, QUANTSLANT earns a commission and you pay the same price. It has no influence on the content, including where the coverage is critical.
More on this from QUANTSLANT
video·Video Rebirth models, access and benchmark guideVideo Rebirth model and access guide with a practical US-English evaluation brief and links to publisher materials.