Retell AI Reaches 100M Monthly Calls With OpenAI Models Behind 80% of Them
Retell AI, a voice AI platform for contact centers, says its monthly call volume has grown from 30 million in early 2026 to more than 100 million by September, with OpenAI models now generating responses for 80% of calls. According to OpenAI's account, Retell has also consumed more than 1 trillion tokens across OpenAI models.
- Retell AI's monthly calls grew from 30 million in early 2026 to more than 100 million by September, across more than 4,000 businesses.
- OpenAI models generate responses for 80% of Retell calls, and customers achieve resolution rates above 80%, according to Retell.
- In Retell's October 13, 2026 benchmark, GPT‑6 Astra ranked first for instruction following at 89.7%.
- Anker reported an 80.4% case resolution rate and an NPS of 63; Pine Park Health reached a 55.7% booking rate.
- Retell has consumed more than 1 trillion tokens across OpenAI models and uses Codex throughout development.
Growth and model choice
Retell AI helps businesses build voice agents that answer questions, take action and hand callers to a person when needed. The platform now serves more than 4,000 businesses. Customers achieve resolution rates above 80%, according to Retell.
The company picks models using its own vCX benchmark, which compares dozens of models on difficult, multi-turn conversations drawn from real production scenarios. It measures instruction following, grounding in the given information, correct tool calls and natural conversational language.
In Retell's October 13, 2026 benchmark, GPT‑6 Astra ranked first for instruction following at 89.7%. Evaluations run before the launch of GPT‑6 Astra showed GPT‑5.5 ranking highest on non-hallucination rate (88.0%) and tool-call correctness (88.3%) with reasoning enabled, ahead of Gemini‑3.1‑Pro and Claude‑4.6 on both measures. It also led on non-hallucination rate with reasoning off.
Retell says it makes new OpenAI models available the day they launch and reruns its benchmark as models improve. Customers can also use OpenAI Realtime models through the Realtime API, and OpenAI text-to-speech as a voice engine.
How a call works and what customers report
Contact center teams define an agent's business context, goals and rules, and developers can configure workflows and integrations. During a call, Retell's harness runs the audio models, tracks the conversation and manages when the agent speaks, listens or handles interruptions. OpenAI models, including GPT‑6.1 Sol, GPT‑6 Astra and GPT‑5.5, generate responses and request tool calls. Retell runs those calls through integrations with the customer's systems and can write outcomes back to a customer record.
When human judgment is needed, the agent briefs the representative, answers follow-up questions and confirms they are ready before connecting the customer. After the call, GPT‑5.6 Terra analyzes the transcript and extracts outcomes and other data.
- Anker: agents handle Tier‑1 post-sales support in the US and UK, including order tracking, warranty checks and troubleshooting. Anker reported an 80.4% case resolution rate and a Net Promoter Score of 63.
- Pine Park Health: agents handle appointment reminders and proactive scheduling and record outcomes in medical charts. The team recovered medical-assistant time equivalent to 2.1 full-time roles, and its scheduling agent achieved a 55.7% booking rate when it reached a patient.
Latency, redundancy and data residency
Daytime traffic on Retell's platform can reach more than 50 times its overnight low. To keep calls responsive at peak demand, Retell uses dedicated servers with OpenAI, shared across its customers.
In the October 13, 2026 benchmark, GPT‑6 Astra and GPT‑6 Sol each recorded median first-sentence latency of 883 milliseconds, while GPT‑5.6 Luna ranked highest at 497 milliseconds. OpenAI's regional API endpoints let Retell route traffic to specific regions, which reduces network travel time and helps meet data residency requirements in industries such as healthcare and financial services.
Retell also runs its own telephony stack with a second stack ready to take over, plus backups for voice and model services.
Codex and GPT‑6 Astra in engineering
Retell's engineers use Codex to research unfamiliar code, build features and debug issues. They give Codex a high-level requirement, and it surveys the code, checks it against actual data and explains what needs to change. For debugging, it cross-references logs and database records with the code. Engineers also use it when working with customers building on OpenAI models, for example to write the instructions that guide a model's responses.
For front-end testing, GPT‑6 Astra's computer use opens the interface when an engineer submits a pull request, clicks through it, tests flows, tries to break them and reports issues. Retell says earlier models could not do this reliably. The company reports $1.6 million in annual revenue run rate per employee and says its engineering team has not needed to grow at the same pace as the business.
What comes next
Retell plans to expand quality assurance and simulation, support more languages and develop knowledge base auto-learning. CEO Bing Wu said the company chooses models "on real calls, not demos." The company's stated goal is to let businesses build, test and improve their own agents, and to have them build an in-house team of internally deployed engineers who own and run it all.