Flat-rate LLM API for AI agents with smart routing and no per-token billing
Foundation model for mobile, wearables, robots, smart home, automotive and microcontrollers
Stage
Not given
Startup
Founded
Not given
Not given
Based in
Not given
Not given
Pricing model
Paid
Open source
Pricing
Starter $19/mo with $30 compute budget (first month), Economy $39/mo with $61 budget (first month), Standard $89/mo with $142 budget (first month), Pro $249/mo with $403 budget (first month), Pro Plus $499/mo with $549 monthly budget, Growth $999/mo with $1,119 budget, Scale $2,499/mo with $2,849 bu
Not given
Key features
Smart routing automatically selects efficient models for routine tasks and frontier models for complex work
Fixed monthly compute budgets with no per-token billing or overage charges
Optional smart pacing to spread monthly budget across the month
Access to frontier models from OpenAI, Anthropic, Google, and other labs
Provider optimization routing to lowest-cost instance of same model
Regional selection with no training on prompts and zero data retention options
Team support with company creation and per-member plan assignment
Dashboard showing real-time compute budget visibility and usage
99.97% API uptime with no session windows or usage limits
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
Predictable flat-rate pricing with no hidden usage windows or surprise cuts
70% further budget stretch through smart routing beyond direct provider pricing
No vendor lock-in with OpenAI-compatible API allowing three-edit switch
Designed for always-on agents with consistent latency and graceful bursty traffic handling
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
Standard Compute