Retell AI, a voice AI platform for contact center operations, now serves more than 4,000 businesses. Monthly call volume grew from 30 million in early 2026 to more than 100 million by September. OpenAI models generate responses for 80% of calls on the platform, customers report resolution rates above 80%, and Retell has surpassed 1 trillion tokens consumed across OpenAI models.

Retell built its vCX benchmark to compare models on difficult multi-turn conversations from real production scenarios, testing instruction following, grounding, tool-call correctness, and conversational language. In its October 13, 2026 benchmark, GPT-6 Astra ranked first for instruction following at 89.7%. Pre-launch evaluations 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, outperforming Gemini-3.1-Pro and Claude-4.6 on both measures; it also led non-hallucination rate with reasoning off. Retell says new customers who choose OpenAI models continue using them. Customers can also use OpenAI Realtime models through the Realtime API and OpenAI text-to-speech as a voice engine. New OpenAI models are made available the day they launch, and the benchmark is rerun as models improve.

During calls, Retell’s harness runs the audio models, tracks conversation, and decides when the agent speaks, listens, or responds to interruptions. OpenAI models including GPT-6.1 Sol, GPT-6 Astra, and GPT-5.5 use customer instructions and conversation context to generate responses and request tool calls; Retell executes those calls through integrations and can write outcomes back to customer records. When human judgment is needed, the agent briefs the representative, answers follow-up questions, and confirms readiness before connecting. After calls, GPT-5.6 Terra analyzes transcripts and extracts outcomes.

Retell reports customer resolution rates above 80%. At 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 an NPS of 63, with tone, pacing, and warmth meeting the expected standard of a human agent. At Pine Park Health, Retell agents handle appointment reminders and proactive scheduling and record outcomes in patients’ medical charts; the team recovered medical-assistant time equivalent to 2.1 full-time roles, and the scheduling agent achieved a 55.7% booking rate when it reached a patient.

Daytime traffic can reach more than 50 times the overnight low. To keep calls responsive, Retell uses dedicated servers with OpenAI shared across customers. In the October 13, 2026 benchmark, GPT-6 Astra and GPT-6 Sol each recorded median first-sentence latency of 883 milliseconds, and GPT-5.6 Luna ranked highest at 497 milliseconds. OpenAI regional API endpoints help Retell reduce network travel time and meet data residency requirements in healthcare and financial services. Retell also builds redundancy into each layer, including its own telephony stack with a second stack ready to take over.

Retell’s engineers use Codex throughout development: they give it a high-level requirement before writing code, let it survey the codebase and check against actual data, and use it to debug by cross-referencing logs and database records. Engineers also use Codex when working with customers building on OpenAI models, for tasks such as writing model instructions. GPT-6 Astra‘s computer use opens and clicks through front-end pull requests, tests flows, tries to break them, and reports issues; earlier models could not do this reliably. Retell says its engineering team has not needed to grow at the same pace as the business, reporting $1.6 million in annual revenue run rate per employee. Next, Retell plans to expand quality assurance and simulation, support more languages, and develop knowledge base auto-learning.