Retell Scales Voice Agent to 100 Million Monthly Calls Using OpenAI Models
THE SHIFT
Can a voice‑assistant handle a hundred million calls a month without breaking the bank? Retell says it now can, after integrating OpenAI’s latest models into its platform.
The company moved from in‑house speech tech to OpenAI’s GPT‑6 Astra for dialogue and Codex for backend logic, promising faster scaling and higher quality interactions.
THE PROOF
OpenAI’s blog notes that Retell’s voice agent benchmark was topped by GPT‑6 Astra, and Codex helped the firm reach a $1.6 M annual revenue run rate per employee.
Retell reports that the combined system now supports roughly 100 M monthly calls, a metric the company uses to illustrate the models’ real‑world throughput.
THE CATCH
The performance gains hinge on OpenAI’s pricing and API availability; cost spikes could erode the reported revenue efficiency.
Benchmark claims are internal, so external validation is pending, and the revenue figure reflects Retell’s specific business model rather than a universal rule.
YOUR MOVE
If you’re building a voice‑first product, monitor Retell’s rollout and consider piloting OpenAI’s dialogue models to test scalability on your own traffic.
Start with a limited‑scope experiment, track cost per interaction, and compare against your current stack before committing to a full migration.
Retell demonstrates that OpenAI models can underpin massive voice‑agent workloads, but the economics depend on API pricing and internal efficiencies.
Sources & further reading
Published by The Breakdown. How we report · Our source desk.
