Retell uses OpenAI models to help power 100M monthly calls

Retell AI, a voice AI platform for contact centers, now handles more than 100 million monthly calls, up from 30 million earlier in 2026, serving over 4,000 businesses. The platform uses OpenAI models to generate responses for 80% of calls, helping maintain call quality and reliability as enterprise workflows grow in complexity. Customers report resolution rates above 80%, with Retell consuming over 1 trillion tokens across OpenAI models to support this scale.

Retell’s vCX benchmark evaluates models on instruction following, grounding, tool use, and conversational language using real production scenarios. In an October 13, 2026 benchmark, GPT‑6 Astra ranked first for instruction following at 89.7%. Before GPT‑6 Astra’s launch, GPT‑5.5 led in non-hallucination rate (88.0%) and tool-call correctness (88.3%) with reasoning enabled, outperforming other models like Gemini‑3.1‑Pro and Claude‑4.6 on these measures.

Contact center teams configure agents with business context, goals, and rules, while Retell’s platform manages audio models, conversation tracking, and tool execution. OpenAI models like GPT‑6.1 Sol, GPT‑6 Astra, and GPT‑5.5 generate responses using customer instructions and conversation context, with Retell executing tool calls through customer system integrations. When human intervention is needed, the voice agent briefs representatives with context and confirms readiness before connecting the caller.

Retell reports $1.6 million in annual revenue run rate per employee, with engineering teams using Codex for code research, feature development, and debugging. The platform supports industries like healthcare and financial services with regional API endpoints to meet data residency requirements. Retell also maintains redundancy across its telephony, voice, and model services to ensure call continuity during service disruptions.