# Avian

> Avian sells dedicated NVIDIA B200 and B300 GPU nodes (complete 8-GPU HGX systems, single-tenant, US region, prepaid reserved terms with the rate locked) and runs a fast, OpenAI-compatible LLM inference API billed per token from $0.1275/M tokens.

## GPU compute

- [GPU compute: B200 and B300 nodes](https://avian.io/compute/): term pricing, node specs, FAQ, how to reserve
- [Machine-readable price feed (JSON)](https://avian.io/compute/prices.json): prices per GPU-hour, terms, stock; public, poll hourly
- NVIDIA B200 (180 GB HBM3e per GPU), per GPU-hour prepaid: 12 months $4.32, 24 months $3.87, 36 months $3.42
- NVIDIA B300 (288 GB HBM3e per GPU), per GPU-hour prepaid: 12 months $4.95, 24 months $4.59, 36 months $4.23
- Sold as complete 8-GPU nodes; multi-node clusters with InfiniBand or RoCE are quoted by sales (info@avian.io).

## Inference API

- [Models](https://avian.io/models/): DeepSeek V4.1, MiniMax M3, GLM-5.3, Kimi K3, MiMo-V2.6 and more
- [Pricing](https://avian.io/pricing/): per-token prices for every model
- [API documentation](https://docs.avian.io/): OpenAI-compatible; works with the OpenAI SDKs

## Company

- [About](https://avian.io/about/)
- [Contact](https://avian.io/contact/): info@avian.io