Documentation
Everything you need to run AI on your own GPU: coding servers for your squad and inference instances for chatbots, OCR, vision and embeddings. From first instance to advanced configuration.
Getting Started
Create your first coding instance and connect your tools in under 10 minutes.
How It Works
Understand syndicAI's architecture, from GPU provisioning to satellite-first data isolation.
API Reference
OpenAI-compatible API documentation, endpoints, authentication, request/response formats, and code examples.
Coding Instances
Coding Instances
What a coding instance is, who it is for and how it works: a shared AI coding server on a dedicated GPU for your squad.
Supported Models
Frontier-class open-source models available on syndicAI, specs, benchmarks, and recommended configurations.
Billing & Pricing
How prepaid credit billing works, weekly GPU budgets explained, squad cost splitting, and spending controls.
Quants & context for coding agents
Quantization formats syndicAI serves on vLLM, FP8, AWQ, MXFP8, NVFP4, W4A16, and more, plus GPU compatibility and context-length guidance for squads.
Inference Instances
Inference Instances
What an inference instance is, how it differs from a coding server, and when to use one.
Inference Sizing and Capacity
How peak requests per minute and a typical request become a GPU choice: memory, speed, parallel requests and the capacity verdict.
Inference Models
The chat, vision, embedding and OCR models you can run on an inference instance, and what each is for.
Inference Limits and Rate Control
Peak requests per minute, per-key request caps, 429 responses, request size limits and what you see when an instance is paused or starting.
Inference API
Call an inference instance: chat completions, embeddings, OCR and image requests, with examples, limits and error responses.
Inference Billing and Availability
How GPU runtime billing works for inference instances, what idle pause does, how restarts behave and what best effort availability means.