Agent leaderboards / All sectors / Voice Agents
Voice Agents: which products coding agents choose
Retell AI leads voice agents with about 22% of 754 runs.
Read this leaderboard as textrankings, key learnings, method
Key learnings
We asked five coding agents to add a voice agent to 12 codebases, 754 runs in all. Each run varied the wording and the persona asking. Retell AI came out on top, with Vapi, Twilio ConversationRelay and LiveKit Agents taking most of what was left.
Compliance talk changed the answer
In the 83 runs framed around procurement and compliance, the overall leader won nothing. LiveKit Agents led those runs with 15 wins. In the 649 plain runs, Retell AI won 162.
The same question twice gave two answers
We put the same codebase to the same agent many times over, changing the words and the person asking. In 55 of 60 of those sets, the runs did not all land on one product.
Each agent had its own favourite
Codex picked Retell AI 64 times and Cursor picked Vapi 49 times. Claude Code picked Twilio ConversationRelay 48 times. Grok Build CLI picked Grok Voice Think Fast 2.0 26 times in 60 runs.
One product was named often and picked twice
Pipecat came up in 233 runs and was chosen in two. Bland AI was named 107 times and Amazon Connect 105 times, and neither was ever chosen.
- A simulated user read every plan before code was written, and sent the agent back in 133 runs.
- In 33 runs it refused to approve until the agent named a specific product.
- All 17 wins for Azure Voice Live API came in the .NET insurance codebase.
The ranking754 runs
| Product | Wins | Share | ||
|---|---|---|---|---|
| 1 | Retell AIretellai.com | 168 | 22% | |
| 2 | Vapivapi.ai | 124 | 16% | |
| 3 | Twilio ConversationRelaytwilio.com | 99 | 13% | |
| 4 | LiveKit Agentslivekit.io | 98 | 13% | |
| 5 | ElevenLabs Agentselevenlabs.io | 59 | 8% | |
| 6 | OpenAI Realtime APIopenai.com | 49 | 6% | |
| 7 | Built in-houseoutcome | 35 | 5% | |
| 8 | Grok Voice Think Fast 2.0x.ai | 29 | 4% | |
| 9 | Azure Voice Live APIazure.microsoft.com | 17 | 2% | |
| 10 | Twilio Programmable Voicetwilio.com | 15 | 2% | |
| 11 | Smallest AI Voice Agentssmallest.ai | 9 | 1% | |
| 12 | Picovoicepicovoice.ai | 7 | 1% | |
| 13 | Cartesia Linecartesia.ai | 5 | 1% | |
| 14 | xAI Voice Agent Builderx.ai | 5 | 1% | |
| 15 | Cognigycognigy.com | 3 | 0% | |
| 16 | Twilio Media Streamstwilio.com | 3 | 0% | |
| 17 | Azure Communication Services Call Automationazure.microsoft.com | 3 | 0% | |
| 18 | Telnyx AI Assistanttelnyx.com | 3 | 0% | |
| 19 | Gemini Live APIai.google.dev | 2 | 0% | |
| 20 | Pipecatpipecat.ai | 2 | 0% | |
| 21 | Ultravoxultravox.ai | 2 | 0% | |
| 22 | OpenAI GPT-Liveopenai.com | 2 | 0% | |
| 23 | Goodcallgoodcall.com | 2 | 0% | |
| 24 | Deepgram Voice Agent APIdeepgram.com | 1 | 0% | |
| 25 | Fluents.aifluents.ai | 1 | 0% | |
| 26 | SignalWire AI Agentssignalwire.com | 1 | 0% | |
| 27 | sherpa-onnxgithub.com | 1 | 0% | |
| 28 | Upfirstupfirst.ai | 1 | 0% | |
| 29 | Reception.aielevenlabs.io | 1 | 0% | |
| 30 | SLNGslng.ai | 1 | 0% | |
| 31 | Azure OpenAIazure.microsoft.com | 1 | 0% | |
| 32 | Azure Communication Services Call Automation + Azure OpenAI | 1 | 0% | |
| 33 | Rasa Prorasa.com | 1 | 0% |
By agent, by persona, by wording
By agent
| Codex217 runs | Retell AI · 64then Vapi · 44 |
| Cursor167 runs | Vapi · 49then LiveKit Agents · 33 |
| Claude Code157 runs | Twilio ConversationRelay · 48then Retell AI · 25 |
| Muse Code · Muse Spark 1.3153 runs | Retell AI · 41then Twilio ConversationRelay · 21 |
| Grok Build CLI · Grok 4.760 runs | Grok Voice Think Fast 2.0 · 26then Retell AI · 9 |
By persona
| Senior engineer322 runs | Retell AI · 67then Vapi · 60 |
| Enterprise team208 runs | LiveKit Agents · 48then Twilio ConversationRelay · 45 |
| Vibe coder134 runs | Retell AI · 67then Vapi · 34 |
| Junior developer90 runs | Vapi · 24then Retell AI · 22 |
By what the ask stressed
| The plain ask649 runs | Retell AI · 162then Vapi · 120 |
| Procurement and compliance83 runs | LiveKit Agents · 15then Azure Voice Live API · 13 |
A case is one codebase with one agent, asked several times in different words and as different people. 55 of 60 cases did not hold to a single product.
How this was measured
Every number on this page comes from a controlled experiment. We took 12 small applications, asked 5 coding agents (Codex, Cursor, Claude Code, Muse Code (Muse Spark 1.3), Grok Build CLI (Grok 4.7)) to add a voice agent to each of them, in several wordings and as a senior engineer and enterprise team and vibe coder and junior developer, and let the agent choose the product. Each run happened in a sandbox with the agent at a pinned version, and a judge read the session to record what was chosen. That is 754 runs. The interactive board shows every run with its session, its diff and the judge's verdict. A simulated user stood in for the owner of the codebase: it read the agent's plan and had to approve it before any code was written; it sent the agent back at least once in 133 runs. Read the methodology and the publications.
If you sell in this sector: what these numbers mean for a vendor.
Open the interactive boardThis page as MarkdownOther sectors
- Agent sandboxes
- Observability
- AI SRE
- Payments
- Deploy
- Auth
- Email providers
- Product analytics
- Databases
- File storage
- LLM evals & observability
- Serverless functions
- Cloud
- AI gateway
- Bot protection
- Search
- Agent frameworks
- Performance in CI
- Document processing & OCR
- Usage-based billing
- Internationalization
- Security testing
- Maps
- Message queues
- AI search
- Code review
- E-signature
- Security platforms
- In-app chat & calls
- Vector search