Smallest.ai

Smallest.ai is a voice AI company based in Bengaluru, India, founded by Sudarshan Kamath and Akshat Mandloi, both previously engineers at Robert Bosch. It builds deliberately small speech models for real-time conversation rather than general-purpose foundation models. The company's thesis is that latency, not raw capability, is what makes a spoken exchange feel human, so it trains compact models that listen, reason and speak concurrently instead of waiting for a complete prompt before answering, handing off to a larger third-party model only when a query genuinely needs one.

The product line spans four first-party models. Lightning is a text-to-speech model the company advertises at roughly 100 milliseconds of latency across more than 15 languages. Pulse is a speech-to-text model covering 38 languages with speaker and emotion detection. Hydra is a native speech-to-speech model, and Electron is a language model under three billion parameters. These sit beneath a voice-agents platform that handles call orchestration, testing and deployment. Smallest.ai states that Electron outperforms GPT-4.1; that claim appears on its marketing site rather than in a published evaluation, so it should be read as a vendor claim rather than an independent result.

In July 2026 the company raised a $13 million Series A, led by Seligman Ventures, with Sierra Ventures and 3one4 Capital participating and total funding passing $21 million. Named customers include RingCentral and Truecaller, both telephony and contact-center businesses that match the real-time voice focus. Chief executive Sudarshan Kamath has said the goal is a voice model a caller cannot distinguish from a person.

Smallest.ai competes in a crowded voice market against ElevenLabs, OpenAI's realtime speech models, and open-weight challengers such as Fish Audio and Voxtral. Its differentiation is architectural rather than scale-based: small models sold on latency and per-minute cost, with published self-serve rates and SOC 2 Type 2, ISO 27001, GDPR and HIPAA compliance claims aimed at regulated contact-center buyers. The trade-off is the usual one for a small vendor in an infrastructure category — the models are cheap and fast, but the company is early, and the frontier-quality claims rest on its own benchmarking.