Foundation model for mobile, wearables, robots, smart home, automotive and microcontrollers
Stage
Startup
Startup
Founded
2024
Not given
Based in
US
Not given
Pricing model
Paid
Open source
Pricing
Pay-as-you-go pricing for serverless inference charged per token and for on-demand deployments charged per GPU usage time. New users receive free credits. Text models range from $0.05 to $1.5 per 1M tokens (input and output), including Llama-3.2-1B-Instruct at $0.06/1M tokens, Meta-Llama-3.1-8B-Inst
Not given
Key features
Text generation with multiple LLM models including Llama, Qwen, and DeepSeek
Embedding and reranking models for RAG applications
Text-to-image and text-to-video generation
Voice generation with ASR/TTS models
Single line of code integration
Agent application with ultra-low latency
Chat2DB application with real-time responsiveness
Inference acceleration capability
Competitive pricing with reduced costs
Tool calling with function selection and argument filling
Structured extraction with guaranteed JSON parsing
Text embeddings for local semantic search and matching
Confidence scoring for routing and gating calls
Laddered subnetwork architecture from 2 to 20 layers
LoRA fine-tuning support for customization
400-4k tokens per second inference speed on Raspberry Pi 5
Python package with pip install and Hugging Face distribution
What makes it different
Out-of-the-box inference acceleration capability
Fast model updates and comprehensive model selection
Single line of code integration with no deployment complexity
Ultra-low latency for responsive applications
Free credits for new users to get started
Beats models 10x its size on mobile tool calling benchmarks
Matches models 2-3x larger on structured extraction tasks
Entire model in a single 8-29 MB binary for tiny devices
Simple Attention Network architecture uses 2x fewer MFLOPs than transformers
Fine-tuned subnetworks outperform larger models like DeepSeek V4 Flash