# Smallest AI

> Smallest AI builds ultra-efficient multimodal AI models and a voice agent platform with sub-400ms latency, supporting text-to-speech, speech-to-text, and speech-to-speech across 30+ languages.

Smallest AI develops small, efficient multimodal AI models and an agentic platform designed to outperform LLMs 100–1000× their size. Their flagship products include Lightning (text-to-speech), Pulse (speech-to-text), Electron (small language model), and Hydra (speech-to-speech), all engineered for ultra-low latency and enterprise-scale deployments. The Atoms platform enables businesses to create, test, deploy, and analyze voice and text agents across channels with minimal setup. Smallest AI targets industries like healthcare, e-commerce, debt collection, logistics, and recruitment.

- **Lightning TTS** — *A text-to-speech model series with 100ms time-to-first-byte, 30+ languages, thousands of accents, voice cloning, and hyper-realistic emotional voices.*
- **Pulse STT** — *A speech-to-text model supporting 38+ languages with code-switching, streaming and batch modes, emotion/speaker/timestamp detection, and interruption handling.*
- **Electron SLM** — *A sub-3B parameter small language model with 45ms TTFT, outperforming GPT-4.1 on multiple benchmarks, specialized for conversational use cases with NSFW and prompt-attack protection.*
- **Hydra S2S** — *A full-duplex multimodal speech-to-speech model with tool calling, asynchronous thinking, and hyper-emotional dialogue.*
- **Atoms Platform** — *A self-learning multi-modal AI agentic platform for building, testing, deploying, and analyzing voice and text agents across voice, email, chat, and social channels.*
- **Developer SDKs** — *Python and Node.js SDKs available for programmatic access to all model APIs.*
- **Enterprise Security** — *SOC 2 Type 2, HIPAA, PCI, GDPR, and ISO-aligned compliance with on-premises deployment options.*
- **On-Premise Deployment** — *Available for enterprise clients requiring data sovereignty and zero data retention.*

## Features
- Text-to-Speech (Lightning) with 100ms TTFB
- Speech-to-Text (Pulse) with 38+ language support
- Small Language Model (Electron) with 45ms TTFT
- Full-duplex Speech-to-Speech (Hydra)
- Voice Cloning
- Atoms agentic platform for voice and text agents
- 30+ language support with local accents and dialects
- Streaming and batch audio processing
- Emotion, speaker, and timestamp detection
- Interruption handling
- Tool calling support
- Asynchronous thinking in Hydra
- Python and Node.js SDKs
- On-premise deployment
- SOC 2, HIPAA, GDPR, PCI compliance
- Sub-400ms average latency
- 99.99% uptime SLA for enterprise

## Integrations
Python SDK, Node.js SDK, REST API, Enterprise SSO, RBAC

## Platforms
LINUX, WEB, API, DEVELOPER_SDK, CLI

## Pricing
Freemium — Free tier available with paid upgrades

## Links
- Website: https://smallest.ai
- Documentation: https://waves-docs.smallest.ai/v4.0.0/content/text-to-speech/quickstart
- Repository: https://github.com/smallest-inc
- EveryDev.ai: https://www.everydev.ai/tools/smallest-ai