Smallest.ai Raises $13M to Make Voice AI Pass the Turing Test

While AI agents are increasingly capable of solving customer support problems, most people can still tell immediately when they're talking to a machine. Smallest.ai, a San Francisco-based startup founded in late 2024, wants to change that — and it just raised $13 million to do it.

What Happened

Smallest.ai closed a $13 million Series A round led by Seligman Ventures, with participation from Sierra Ventures and 3one4 Capital. The fresh capital brings total funding to over $21 million, according to TechCrunch. The startup builds small, specialized voice models designed to handle real-time conversation — not by making LLMs faster, but by rethinking how voice AI processes speech entirely.

The core insight: humans listen, think, and speak simultaneously. Most AI voice pipelines don't. Smallest.ai's model mimics this parallel processing, achieving sub-second latency by running its proprietary Hydra speech-to-speech model as a real-time intelligence layer.

Why It Matters

Founder and CEO Sudarshan Kamath believes all AI agents will soon rely on two models: a small voice model for real-time interaction, and an "offline" LLM called upon only for complex problems. When Smallest.ai's model encounters a subject outside its knowledge base, it hands off to a larger model — briefly placing the customer on hold to "research" the issue, just as a human would.

This architecture targets a growing market: enterprise voice agents for customer support. Existing customers include RingCentral and Truecaller. The model handles accents, operates in noisy environments, and supports 38 languages with features like emotion detection and automated PII redaction.

The company competes with voice AI leader ElevenLabs, as well as Cartesia and regional players like Sarvam. But unlike competitors applying voice AI to podcasting and dubbing, Smallest.ai focuses exclusively on real-time conversational agents.

Builder Impact: For AI developers building customer-facing agents, Smallest.ai's two-model architecture offers a practical blueprint — pair a lightweight real-time voice model with a heavy reasoning backend. This pattern could become standard for any latency-sensitive agent deployment, from support bots to in-car assistants.