Key facts
- Smallest.ai has raised $13 million in Series A funding.
- The startup aims to create voice AI that is indistinguishable from human conversation.
- Their model processes information by listening, thinking, and speaking simultaneously.
- The company uses smaller, specialized models for real-time interaction, handing off complex queries to larger models.
- Existing customers include RingCentral and Truecaller.
Smallest.ai, a startup founded in late 2024, has raised $13 million in a Series A funding round to develop voice AI that aims to be indistinguishable from human conversation. The company believes the next advancement in voice agents will come from smaller, specialized models rather than faster large language models.
Led by Seligman Ventures with participation from Sierra Ventures and 3one4 Capital, the funding brings Smallest.ai's total capital to over $21 million. Founder and CEO Sudarshan Kamath explained that their model mimics human conversation by processing information in real-time, allowing for simultaneous listening, thinking, and speaking, thus minimizing response lag.
When a query falls outside the specialized model's scope, it is handed off to a larger foundational model, with the customer briefly placed on hold, similar to how a human would handle such a situation. Kamath envisions a future where AI agents utilize a combination of a small, real-time voice model and an on-demand larger LLM for complex tasks.
Smallest.ai's models are designed to handle voice-specific nuances, including diverse accents, multiple languages, and operation in noisy environments. The startup's current clients include companies in the voice sector like RingCentral and Truecaller, with potential customers including any customer support company. Kamath stated that focusing on voice AI is a core competency for Smallest.ai, preventing distractions for customer support startups.