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
100+ models via API (DeepSeek, Qwen, Llama, Nemotron, Gemini, Step, Cosmos)
Zero data retention policy
SOC 2 and ISO 27001 certified
Pay-as-you-go, per-token pricing
Cached input pricing (up to 80% discount)
Up to 1M-token context windows (DeepSeek-V4, GLM-5.2, Nemotron-3-Ultra)
Inference-optimized US data centers
Text, image, speech, video, world model inference APIs
Private deployments with dedicated support
On-demand DGX B300 GPU rental ($4.20/instance-hour)
OpenAI SDK compatible API
Step 3.7 Flash (198B MoE vision-language model)
NVIDIA Cosmos 3 World Model for physical AI
Automatic speech recognition (ASR) models
Text-to-music, text-to-video, world model inference
