Retell runs 100M monthly calls on OpenAI models
Voice AI platform Retell says OpenAI models now generate responses for 80% of its more than 100 million monthly contact-center calls, an enterprise voice deployment at unusual scale.
Retell AI, a voice agent platform for contact centers, has passed 1 trillion tokens consumed across OpenAI models, according to a customer case study published by OpenAI. Retell serves more than 4,000 businesses. Its monthly call volume grew from 30 million in early 2026 to more than 100 million by September, and it reports customer resolution rates above 80%.
During a call, Retell's harness runs the audio models and handles turn-taking and interruptions. OpenAI models, including GPT‑6.1 Sol, GPT‑6 Astra and GPT‑5.5, generate responses and request tool calls, which Retell then executes against customer systems. After each call, GPT‑5.6 Terra analyzes the transcript.
Retell built its own benchmark, vCX, from multi-turn production scenarios. In its October 13 run, GPT‑6 Astra ranked first for instruction following at 89.7%. Earlier evaluations had GPT‑5.5 ahead of Gemini‑3.1‑Pro and Claude‑4.6 on non-hallucination rate (88.0%) and tool-call correctness (88.3%) with reasoning enabled. For median first-sentence latency, GPT‑5.6 Luna led at 497 ms, while GPT‑6 Astra and GPT‑6 Sol each recorded 883 ms. Retell designed the benchmark and runs it, and it has not been independently verified.
Two customers are named. Anker reported an 80.4% case resolution rate and an NPS of 63 for Tier‑1 post-sales support. Pine Park Health said it recovered medical-assistant time equal to 2.1 full-time roles.
Internally, Retell engineers use Codex for code research, implementation and debugging, and use GPT‑6 Astra's computer use to test front-end pull requests. Retell says this helps it reach an annual revenue run rate of $1.6M per employee. The case study does not give total revenue or headcount.