The round arrives as the voice-AI market heats up and as investors hunt for the next leap beyond text-based chatbots. For continuous coverage of the startups reshaping the artificial-intelligence industry, follow our latest AI news.
A Bet on Small, Specialized Voice Models
Smallest.ai's core thesis is that the next breakthrough in voice agents will not come from making large language models faster, but from building smaller, purpose-built models engineered specifically for human conversation. Chief executive Sudarshan Kamath told TechCrunch that today's LLMs are fundamentally mismatched to the rhythm of speech: they wait for a complete prompt before they begin "thinking," a delay that feels natural in a text chat but jarring on a phone call.
"While I'm speaking to you, you're already thinking, and you might interrupt me if I talk for too long," Kamath explained, describing the real-time, overlapping nature of human dialogue that his startup's model is designed to replicate. The model processes audio by listening, thinking, and speaking simultaneously, aiming for virtually zero response lag.
A Two-Model Architecture
Smallest.ai's system relies on what Kamath predicts will become a standard architecture for AI agents: a small, lightning-fast voice model handles the live conversation, while an "offline" large foundational model is summoned only when a query exceeds the smaller model's knowledge base. When that happens, the system briefly places the caller on hold to "research" the issue, much as a human agent might.
That division of labor lets the startup focus its small model strictly on voice-specific challenges, including handling diverse accents, supporting dozens of languages, and operating in noisy environments. By contrast, competitors that apply general-purpose LLMs to voice must contend with far higher latency and compute costs.
Customers and Competition
Smallest.ai's existing customers include companies in the voice and communications space, among them RingCentral and Truecaller. Kamath argued that well-funded customer-support startups have little incentive to build their own voice models, since becoming "extremely good at doing voice is a distraction from their core business."
The startup competes with voice-AI leader ElevenLabs, as well as Cartesia and regional players such as Sarvam that focus on local languages. While some rivals apply voice AI to dubbing and podcasting, Smallest.ai concentrates exclusively on real-time conversational agents for enterprise customers.
The Turing Test for Voice
Kamath's ambition is unambiguous. "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."
Whether that bar is achievable in the near term remains an open question. Voice AI has improved dramatically in naturalness over the past two years, but even the best systems still struggle with the subtle cues — hesitation, humor, emotional inflection — that make human conversation feel alive. Smallest.ai's bet is that a model purpose-built for speech, rather than adapted from a text-trained LLM, can close that gap faster than the industry expects.
A Crowded but Fast-Growing Market
The funding underscores broader investor enthusiasm for the voice-AI category. Voice interfaces are increasingly seen as the most natural way for consumers to interact with AI agents, and enterprise call centers represent a massive, immediately addressable market where even modest improvements in latency and naturalness can translate into measurable cost savings.
With $21 million now in hand, Smallest.ai plans to expand its model capabilities and grow its enterprise customer base. The competitive landscape, however, is unforgiving: ElevenLabs commands significant market share, and larger players continue to integrate voice into their flagship products. Smallest.ai's differentiation hinges on the premise that a focused, specialized voice model can outperform generalist systems on the one metric that matters most for real-time conversation — speed that feels human.
Stay Ahead of AI
From voice agents to frontier models, the deals and breakthroughs are coming fast. Bookmark AI Buzz Wire for ongoing coverage of the companies and founders building the future of artificial intelligence.
Read more AI startup news →
