Smallest.ai Secures $13M to Narrow the Gap Between AI and Human Speech

The Series A round, led by Seligman Ventures with participation from Sierra Ventures and 3one4 Capital, brings the company’s total funding to over $21 million. Founder and CEO Sudarshan Kamath argues that standard large language models (LLMs) are hindered by the latency inherent in processing entire prompts before generating a response. To eliminate this, Smallest.ai is developing a specialized intelligence layer that listens, thinks, and speaks simultaneously, mimicking the interruptibility and responsiveness of human dialogue.

Smallest.ai operates on a dual-model architecture. A compact, voice-tuned model handles real-time interaction, while an offline LLM is triggered only when complex problem-solving is required—a process Kamath likens to a human agent placing a caller on hold to research an issue. By focusing exclusively on voice-specific challenges like background noise, accents, and multilingual support, the startup aims to provide a plug-and-play solution for enterprise customers like RingCentral and Truecaller. Kamath believes that as AI agents become ubiquitous, companies will prefer integrating dedicated voice specialists rather than diverting resources to build their own proprietary models.