Features
On-device inference with sub-120ms latency
Hybrid cloud routing based on model confidence
Automatic cloud fallback for complex/noisy requests
Transcription with <6% WER and privacy mode
Tool calling and function calling (Needle 26M model)
Voice activity detection (Silero VAD)
Multi-platform SDK (iOS, Android, macOS, wearables)
INT4/INT8 quantization with zero-copy memory mapping
NPU acceleration on Apple, Snapdragon, Exynos, MediaTek
OpenAI-compatible API endpoints
Cactus Graph for custom model implementation
Cactus Kernels: custom attention with KV-cache quantization
TurboQuant-H: 2-bit embedding quantization for Gemma 4
Needle 26M distilled model for high-speed tool calling
Offline-capable inference mode
Pre-recorded speech-to-text (Universal-3.5 Pro, 18 languages)
Pre-recorded speech-to-text (Universal-2, 99 languages)
Real-time streaming speech-to-text (Universal-3.5 Pro Realtime with Context Carryover)
Voice Agent API with built-in turn detection and interruption handling
Speech Understanding API (speaker ID, sentiment, chapters, summaries)
Guardrails for inline PII redaction and content moderation
LLM Gateway routing across GPT, Claude, and Gemini with fallback
Keyterms Prompting to improve accuracy for specific terms
Agent Management API to store and manage agent configurations
HTTP Tool Calling for Voice Agent API (no proxy server needed)
Production-ready Python and TypeScript SDKs
Self-hosted Voice AI Cloud deployment for enterprise
No concurrency limits, no throttles, no forced commitments
Global redundancy and enterprise-grade uptime
Code-switching support (Universal-3.5 Pro)