Smallest.ai, a startup founded in late 2024, has raised $13 million in Series A funding to develop ultra-fast voice AI models designed to make conversations with AI agents indistinguishable from human interactions. The round was led by Seligman Ventures, with participation from Sierra Ventures and 3one4 Capital. After the Series A round, the company's total funding stands at over $21 million.
Most people can still tell immediately when talking to a machine instead of a human. Smallest.ai wants to change that. The company is developing a small voice model designed to mimic human processing by listening, thinking, and speaking simultaneously. The model serves as a real-time intelligence layer for natural customer conversations with virtually zero response lag.
Why Smaller Models Matter for Voice
Large language models work well in text chat, but they stumble in spoken conversation. LLMs work by taking an entire prompt and then thinking, which is acceptable in text chat but not voice. In voice conversation, even a short pause feels unnatural.
Sudarshan Kamath, founder and CEO of Smallest.ai, explained the fundamental difference. "While I'm speaking to you, you're already thinking, and you might interrupt me if I talk for too long," he said. "The way an LLM works is you give it an entire prompt, and then it starts thinking."
Kamath continued: "If you think about how we are talking, I'm not giving you like a large clipping of my audio, and then you start thinking."
That distinction drives the company's entire approach. The next leap in voice agents will come from smaller, specialized models, not faster LLMs.
Handling the Unknown
No model knows everything. Smallest.ai has a plan for that. If the model encounters a subject outside its knowledge base, it hands off to a large foundational model, briefly placing the customer on hold.
That handoff keeps conversations moving without forcing the small model to guess. The design allows the system to stay fast in normal exchanges while still accessing deeper knowledge when needed.
Kamath believes all AI agents will soon rely on two models: a small voice model for real-time interaction and an offline LLM for complex problems. That two-model architecture, he argues, is the future of conversational AI.
Focus on Voice Nuances
Smallest.ai focuses on voice-specific nuances: diverse accents, dozens of languages, noisy environments. These are areas where large foundational models often fall short. The company builds specifically for real-time conversational voice agents for enterprise customers, not audio dubbing or podcasting.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
That focus sets it apart from competitors like ElevenLabs and Cartesia, which apply voice AI to a broader range of use cases. Regional players like Sarvam focus on local languages. Smallest.ai stays strictly in the enterprise voice agent lane.
Existing customers include RingCentral and Truecaller. Newer AI customer support companies Sierra and Decagon are mentioned as potential customers.
Customer Support Startups Should Focus
Kamath has a clear view on where customer support startups should spend their energy. He said customer support startups becoming "extremely good at doing voice is a distraction from their core business."
That comment reflects the company's positioning. Smallest.ai wants to be the voice layer that other companies plug in, rather than forcing each startup to build its own voice technology from scratch.
The company's ambition goes beyond practical business goals. Kamath stated the goal is to break the Turing test. "We want our models to break the Turing test," he said. "You should speak to our model and not know it's AI or human. That's the sole focus of the company."
The Road Ahead
Smallest.ai was founded in late 2024, making it a young company with a fast trajectory. The $13 million Series A round and the total funding of over $21 million give it room to scale its voice models and expand its enterprise customer base.
The company's approach is distinct from simply making LLMs faster. By building a small model that handles real-time interaction and handing off complex problems to larger models, Smallest.ai is betting that the future of voice AI lies in specialization, not size.
Whether the model can truly break the Turing test remains to be seen. But the company's early traction with customers like RingCentral and Truecaller suggests there is demand for voice AI that feels natural in real conversations.
For now, Smallest.ai is focused on one thing: making machines sound human. The company believes that goal is within reach, and its investors are betting $13 million that it is.