Smallest.ai raises $13M for voice AI
Smallest.ai closed a $13 million Series A for its real-time, asynchronous voice AI, bringing total funding past $21 million as it targets contact centers.
Smallest.ai, a voice-AI startup founded in late 2024, closed a $13 million Series A with Sierra Ventures and 3one4 Capital among the investors, bringing its total raised to more than $21 million. The company builds specialized voice models for real-time conversational agents used in enterprise customer support, with existing customers including RingCentral and Truecaller.
The pitch is architectural rather than purely model-quality: Smallest.ai’s models listen, think, and respond simultaneously instead of waiting for a full utterance to end, mimicking how humans naturally interrupt and overlap in conversation. Founder and CEO Sudarshan Kamath described the design goal this way:
While I’m speaking to you, you’re already thinking, and you might interrupt me if I talk for too long.
When a query needs more reasoning than the voice model can supply, the system hands off to a larger language model while keeping the caller briefly on hold, rather than dropping to a generic, disjointed fallback. The company also emphasizes accent handling, multilingual support across dozens of languages, and reliable operation in noisy environments, competing against ElevenLabs, Cartesia, and regional players like Sarvam.
The round is a small one by 2026 AI standards, but it lands in a voice-AI market where response latency and turn-taking naturalness, not just transcription accuracy, have become the visible product differentiator for anyone deploying a phone-facing agent.
If you’re evaluating voice agents for support or sales workflows, test vendors on interruption handling and hold-time behavior under real call conditions, the failure mode that turns an otherwise capable agent into an obviously robotic one, rather than relying on a benchmark built around clean, non-overlapping speech. It’s the same crowded, latency-obsessed field Gradium raised $100M into and xAI is building a voice agent platform for; the differentiator across all of them is turn-taking behavior, not raw transcription quality.