Retell Scales Voice AI to 100 Million Calls Using OpenAI’s GPT‑6 Astra

Retell, a specialist in real‑time voice agents, announced that its platform now processes roughly 100 million calls each month. The milestone reflects a deep integration of OpenAI’s newest large language model, GPT‑6 Astra, which topped Retell’s internal voice‑agent benchmark for latency, accuracy, and naturalness.

Beyond the model itself, Retell leverages OpenAI’s Codex system to automate large portions of its development pipeline. This automation has helped the company achieve an annualized revenue run‑rate of about $1.6 million per employee, a figure that underscores the productivity gains possible when cutting‑edge AI tooling meets a focused product strategy.

Why it matters for GPU and AI infrastructure

Running GPT‑6 Astra at the scale required for 100 million concurrent voice interactions demands massive, low‑latency inference capacity. Cloud providers that can deliver high‑density GPU clusters with optimized networking will be essential partners for companies like Retell as they push real‑time conversational AI into mainstream telephony and customer‑service workloads.

As voice‑AI adoption accelerates, the industry will see tighter coupling between model performance, inference economics, and specialized hardware. Retell’s trajectory illustrates how advances in model architecture and developer tooling translate directly into infrastructure demand, shaping procurement decisions for GPU‑rich platforms worldwide.