How many businesses use Retell AI?

Retell's platform serves more than 4,000 businesses, with monthly call volume growing from 30 million in early 2026 to over 100 million by September.

Which OpenAI models does Retell use for voice agents?

Retell uses OpenAI models including GPT-6.1 Sol, GPT-6 Astra, and GPT-5.5 to generate responses during calls, and GPT-5.6 Terra to analyze transcripts after calls.

How does Retell measure model performance?

Retell built its vCX benchmark to compare models on multi-turn conversations from real production scenarios. In its October 13, 2026 benchmark, GPT-6 Astra ranked first for instruction following at 89.7%.

Large Language ModelsAudio & SpeechAI Business & IndustryOpenAI Blog3m agoOct 28, 2026, 10:00 JST

Retell AI hits 100 million monthly calls on OpenAI models

Send on LINE

3 Key Points

What happened
Retell AI, a voice AI platform for contact centers, grew monthly call volume from 30 million in early 2026 to more than 100 million by September, serving over 4,000 businesses. OpenAI models now generate responses for 80% of its calls, and Retell has surpassed 1 trillion tokens consumed across OpenAI models.
Why it matters
The scale of Retell's adoption, with 80% of calls handled by OpenAI models and resolution rates above 80%, suggests enterprise contact centers are increasingly comfortable letting AI agents handle live customer conversations, not just demos.

WHO IT HITSEnterprise contact center operators and the teams that manage customer support workflows are the clearest audience. The fact that Retell's agents resolve more than 80% of requests suggests these operators may need to rethink staffing and escalation paths as AI handles a larger share of live calls.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Get these free every morning

Context & Analysis

Retell's growth from 30 million to over 100 million monthly calls in roughly nine months shows that voice AI platforms are moving from pilot programs into production workloads. The company's decision to standardize on OpenAI models for the majority of calls reflects a bet that model quality and reliability matter more than flexibility at scale.

Retell's vCX benchmark, which tests instruction following, grounding, tool calls, and conversational language, gives the company a way to justify model choices with data from real calls rather than vendor demos. That approach appears to resonate with enterprise customers who need agents to follow company policies and stay within assigned roles during unpredictable conversations.

The operational details, such as dedicated servers shared across customers and regional API endpoints, suggest that latency and data residency are becoming practical requirements for voice AI deployments in regulated industries. Retell's reported $1.6 million in annual revenue run rate per employee, alongside a hiring pace that trails business growth, hints at how AI-assisted engineering may change the cost structure of scaling a software company.

FAQ

How many businesses use Retell AI?: Retell's platform serves more than 4,000 businesses, with monthly call volume growing from 30 million in early 2026 to over 100 million by September.
Which OpenAI models does Retell use for voice agents?: Retell uses OpenAI models including GPT-6.1 Sol, GPT-6 Astra, and GPT-5.5 to generate responses during calls, and GPT-5.6 Terra to analyze transcripts after calls.
How does Retell measure model performance?: Retell built its vCX benchmark to compare models on multi-turn conversations from real production scenarios. In its October 13, 2026 benchmark, GPT-6 Astra ranked first for instruction following at 89.7%.

OpenAI BlogRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.