Smallest.ai Raises $13M Series A to Build Ultra-Fast Voice AI That Mimics Human Conversation

Smallest.ai Closes $13 Million Series A Funding Round

Smallest.ai has raised $13 million in Series A funding led by Seligman Ventures, with participation from Sierra Ventures and 3one4 Capital1

1

2

2

. This brings the startup's total funding to over $21 million since its founding in late 20241

1

. The fresh capital will accelerate development of voice AI models designed to make conversations with AI agents indistinguishable from speaking with humans. Founded by Sudarshan Kamath, Smallest.ai is betting that the next leap in voice AI won't come from making large language models faster, but from using smaller, specialized models built specifically for human conversation1

1

.

Introducing Hydra and Voice 4.0 Architecture for Real-Time Conversational Flows

Coinciding with the funding announcement, Smallest.ai unveiled Hydra, an asynchronous speech-to-speech model built on its most advanced Voice 4.0 architecture2

2

. This represents a fundamental shift in how voice AI systems operate. Traditional voice AI relies on a chained stack of separate technologies including speech recognition, large language model processing, orchestration layers, memory systems, text-to-speech engines, and guardrails that must execute sequentially2

2

. This disjointed process creates noticeable latency and interactions that feel unmistakably artificial. Hydra's asynchronous voice AI architecture processes listening, reasoning, taking actions, and responding in parallel rather than sequentially2

2

. "Humans don't wait for someone to finish speaking before they begin thinking. We listen, think, and respond simultaneously," Kamath explained. "Voice AI needs to work the same way"2

2

.

How Smallest.ai Achieves Human-Like Interactions Through Specialized Voice AI Models

Smallest.ai's approach fundamentally differs from conventional AI agents. The startup's model serves as a real-time intelligence layer that enables natural customer conversations on specific topics with virtually zero response lag1

1

. When the model encounters a subject outside its limited knowledge base, it hands off the query to a large foundational model, briefly placing the customer on hold to "research" the issue, just as a real human would do1

1

. "While I'm speaking to you, you're already thinking, and you might interrupt me if I talk for too long," Kamath told TechCrunch, explaining how the startup's model mimics human information processing1

1

. Kamath believes all AI agents will soon rely on two models: a small voice model for real-time interaction and an "offline" LLM called upon as needed to solve complex problems1

1

.

Tackling Latency and Voice-Specific Challenges in Enterprise Customer Support

Unlike large foundational models, Smallest.ai focuses strictly on voice-specific nuances including handling diverse accents, supporting dozens of languages, and operating in noisy environments1

1

. The company addresses a critical problem: while latency is acceptable in text chat, even a short pause in voice conversation feels unnatural1

1

. Smallest.ai has developed speech-to-text models such as Pulse STT Pro and Lightning V3.1, which consistently rank among the highest voice AI systems on the Artificial Analysis benchmark2

2

. The original Lightning model, released last year, was described as the fastest text-to-speech model on the market, able to generate 10 seconds of speech in 100 milliseconds2

2

. Lightning has since expanded to support 38 languages and been enriched with emotion detection, speaker diarization, data redaction, and noise reduction features2

2

.

Current Customers and Market Positioning Against ElevenLabs and Cartesia

Smallest.ai's existing customers include companies in the voice space such as RingCentral and Truecaller, along with Kogtal Financial and Readymode1

1

2

2

. These deployments have helped reduce customer support costs by as much as 80% in some cases2

2

. Kamath said any customer support company, including newer ones like Sierra and Decagon, represents a potential customer for the startup1

1

. When asked why well-funded AI customer support companies wouldn't build their own voice models, Kamath argued that for customer support startups, becoming "extremely good at doing voice is a distraction from their core business"1

1

. Smallest.ai competes with voice AI leader ElevenLabs, as well as Cartesia and regional players like Sarvam that focus on local languages1

1

. While some competitors apply voice AI to use cases like audio dubbing and podcasting, Smallest.ai focuses strictly on real-time conversational voice agents for enterprise customers1

1

.

The Growing Voice AI Market and Quest to Pass the Turing Test

The global voice AI industry is currently valued at just $2.4 billion annually but is expected to grow to more than $47.5 billion by the end of 2034, according to a study by Market.US2

2

. Despite this enormous potential, AI currently accounts for less than 1% of the world's voice interactions2

2

. Existing voice AI systems simply aren't able to handle the complexity of real-world interactions, which is why they're only deployed in very narrow customer service use cases2

2

. "We want our models to break the Turing test," Kamath stated. "You should speak to our model and not know it's AI or human. That's the sole focus of the company"1

1

. Watch for Smallest.ai to expand its language support and deploy Hydra across more enterprise customer support scenarios as it pursues this ambitious goal of creating truly indistinguishable human-like interactions.