Smallest.ai Revolutionizes Voice AI with $13 Million Raised

⚡

Key Takeaways

1Smallest.ai has raised $13 million to develop a voice AI indistinguishable from a human, focusing on specialized models.

2The startup uses a vocal model that mimics human information processing, enabling natural and latency-free interactions.

3In the case of complex requests, Smallest.ai switches to a large foundational model, thus providing a hybrid solution for customer support.

💡Why it matters — This innovation could transform the customer support sector by making AI interactions smoother and more natural, thereby reducing costs and enhancing the user experience.

Smallest.ai Bets on a Voice AI Indistinguishable from Humans

In the ever-evolving world of artificial intelligence agents, the ability to solve customer support issues is continuously improving. However, a clear distinction between a machine and a human still persists for most users.

Founded in late 2024, Smallest.ai is a startup that is banking on a major technological leap for voice agents. Unlike the conventional approach that prioritizes the acceleration of large language models, the company focuses on more compact and specialized models, specifically designed for human conversation. The goal is to make interactions with these AI agents as natural as those with a human interlocutor.

To achieve this goal, Smallest.ai is developing a compact voice model that mimics the human process of listening, thinking, and speaking simultaneously. Sudarshan Kamath, the founder and CEO of the company, explains that this model is designed to think in real-time, allowing for natural interruptions, just as a human would during a conversation.

$13 Million in Funding to Support Innovation

To bring this vision to life, Smallest.ai recently raised $13 million in a Series A funding round. This round was led by Seligman Ventures, with participation from Sierra Ventures and 3one4 Capital. With this new influx of capital, the total funding for the startup now exceeds $21 million.

Kamath emphasizes that, unlike a large language model that requires a complete prompt before it starts to think, Smallest.ai's model operates in real-time. This responsiveness is crucial in voice conversations where even a brief pause can seem artificial.

A Hybrid Approach for More Natural Interactions

The model developed by Smallest.ai acts as a layer of real-time intelligence, enabling smooth customer conversations on specific topics without noticeable delays. However, when the model encounters a topic outside its knowledge base, the request is transferred to a large foundational model. This hybrid approach allows for briefly placing the customer on hold to "search" for the issue, thereby mimicking the behavior of a human agent.

Kamath is convinced that the future of AI agents lies in the use of two distinct models: a voice model for immediate interaction and a large language model for more complex queries.

A Focus on Voice Specificities

Unlike large language models, Smallest.ai focuses on the nuances of voice, such as managing accents, supporting multiple languages, and performing well in noisy environments. This specialization allows the startup to stand out in the industry.

Among Smallest.ai's current clients are voice sector companies like RingCentral and Truecaller. Kamath believes that any customer support company, including newcomers like Sierra and Decagon, could benefit from their technology.

A Competitive but Promising Market

Facing competitors such as ElevenLabs, Cartesia, and regional players like Sarvam, Smallest.ai distinguishes itself by focusing on real-time conversational voice agents for businesses. While others apply voice AI to areas like audio dubbing and podcasting, Smallest.ai remains focused on its primary goal.

"We want our models to pass the Turing test," says Kamath. The company's ambition is clear: to enable users to converse with their models without being able to distinguish whether they are speaking to a machine or a human.