Smallest.ai Speech AI Software

Smallest.ai focuses on specialized, compact voice models intended to make real-time agent conversations feel human, with low lag and support for accents, languages, and noisy environments. Rather than relying only on large cascaded stacks, the company markets native speech-to-speech and small models that listen, think, and speak with minimal delay, handing complex research to larger models when needed. Product surfaces include APIs for speech generation and recognition, an agent playground, and enterprise options such as on-prem or private cloud deployment. Models highlighted on the site include Lightning for text to speech, Pulse for speech to text, Electron as a small language model, and Hydra for speech to speech. Founders Sudarshan Kamath and Akshat Mandloi are named on the company press kit. TechCrunch covered a thirteen million dollar Series A closed around July 2026, led by Seligman Ventures with Sierra Ventures and 3one4 Capital, bringing total funding above twenty-one million dollars, and identified RingCentral and Truecaller among existing customers. The company press release and site place headquarters in San Francisco at 311 California Street, Suite 320. Pricing pages publish pay-as-you-go agent rates that vary by model architecture, typically in a range around nine to twenty-one cents per minute on standard configurations, with enterprise discounts and dedicated support available. Exact employee counts and revenue are not disclosed on the primary sources used for this entry.

1. Lightning V3.1 Pro text-to-speech marketed at about 100 ms latency across 15+ languages

2. Pulse Realtime speech-to-text with multilingual transcription, emotion, and speaker detection

3. Electron sub-3B small language model for voice agent reasoning

4. Hydra native speech-to-speech model for production conversations

5. Agentic platform to configure voices, languages, and go-live workflows

6. Developer APIs for rapid prototyping and integration

7. Enterprise security claims including ISO 27001, SOC 2 Type 2, GDPR, and HIPAA

8. Deployment options spanning cloud and on-prem or private infrastructure

Published on smallest.ai/pricing. Pay as you go starts with $10 free credits and full API/model access with no long-term contracts. Agent pricing depends on cascading (hosting + STT + LLM + TTS + telephony) or speech-to-speech stacks, listed about $0.09 to $0.21 per minute depending on models, with enterprise as low as $0.05 per minute. Hosting fee $0.01/min flat on pay as you go. Example model layer rates include STT about $0.009/min, TTS about $0.09/min, and various LLM realtime options at listed per-minute amounts. Phone numbers $10/number/month. Add-ons include knowledge base, PII removal, and HIPAA zero data retention at $1,000/mo on pay as you go. Enterprise includes custom pricing, SLAs, SSO/compliance, on-prem, and forward-deployed support.

RingCentral; Truecaller