# OpenRelay

> The inference delivery network (IDN): inference routed across pooled GPU
> capacity, and on-demand cloud GPUs billed hourly with no commitment, quota,
> or reservation. Deploy a GPU VM in minutes, or point an existing OpenAI
> integration at our gateway with a one-line base URL change.
> Backed by Y Combinator (S26). Formerly VectorLay.

## GPU pricing (list rate, per GPU-hour, current)

- NVIDIA A100 80GB (80GB HBM2e): $1.10/hr
- NVIDIA B200 (180GB HBM3e): $6.30/hr
- NVIDIA B300 (288GB HBM3e): $8.49/hr
- NVIDIA H100 SXM (80GB HBM3): $2.60/hr
- AMD Instinct MI355X (288GB HBM3E): $6.30/hr
- NVIDIA RTX 3090 (24GB GDDR6X): $0.18/hr
- NVIDIA RTX 4090 (24GB GDDR6X): $0.35/hr
- NVIDIA RTX 5090 (32GB GDDR7): $0.50/hr
- NVIDIA RTX Pro 6000 (96GB GDDR7): $1.55/hr

Multi-GPU VMs multiply the per-GPU rate. Full catalog and configurator: https://openrelay.inc/gpus

## Inference pricing (USD per 1M tokens, current)

Three tiers, priced off one catalog rate:

- Sync: real-time requests at the catalog rate.
- Async: queued requests, results delivered later, at 20% off Sync. Early access, not on the public API yet.
- Batch: a JSONL file via /v1/batches, completed within 24 hours, at 50% off Sync. Runs on the models marked below.

| Model | Tier | Input | Cached input | Output |
|---|---|---|---|---|
| openrelay/gpt-oss-120b | Sync | $0.15 | $0.015 | $0.60 |
| openrelay/gpt-oss-120b | Async | $0.12 | $0.012 | $0.48 |
| openrelay/gpt-oss-120b | Batch | not offered | not offered | not offered |
| openrelay/gpt-oss-20b | Sync | $0.05 | $0.005 | $0.20 |
| openrelay/gpt-oss-20b | Async | $0.04 | $0.004 | $0.16 |
| openrelay/gpt-oss-20b | Batch | not offered | not offered | not offered |
| openrelay/gemma-4-31b | Sync | $0.99 | $0.99 | $1.49 |
| openrelay/gemma-4-31b | Async | $0.792 | $0.792 | $1.192 |
| openrelay/gemma-4-31b | Batch | $0.495 | $0.495 | $0.745 |
| openrelay/glm-5.2 | Sync | $1.82 | $0.338 | $5.72 |
| openrelay/glm-5.2 | Async | $1.456 | $0.2704 | $4.576 |
| openrelay/glm-5.2 | Batch | not offered | not offered | not offered |
| openrelay/glm-5.3 | Sync | $0.079 | $0.062 | $3.00 |
| openrelay/glm-5.3 | Async | $0.0632 | $0.0496 | $2.40 |
| openrelay/glm-5.3 | Batch | not offered | not offered | not offered |
| openrelay/glm-5.3-flash | Sync | $0.045 | $0.009 | $0.45 |
| openrelay/glm-5.3-flash | Async | $0.036 | $0.0072 | $0.36 |
| openrelay/glm-5.3-flash | Batch | not offered | not offered | not offered |
| openrelay/qwen3.8-27b | Sync | $0.42 | $0.085 | $3.00 |
| openrelay/qwen3.8-27b | Async | $0.336 | $0.068 | $2.40 |
| openrelay/qwen3.8-27b | Batch | not offered | not offered | not offered |
| openrelay/deepseek-v4.1-flash | Sync | $0.15 | $0.006 | $1.10 |
| openrelay/deepseek-v4.1-flash | Async | $0.12 | $0.0048 | $0.88 |
| openrelay/deepseek-v4.1-flash | Batch | not offered | not offered | not offered |
| openrelay/deepseek-ocr-2 | Sync | $0.039 | n/a | $0.039 |
| openrelay/deepseek-ocr-2 | Async | $0.0312 | n/a | $0.0312 |
| openrelay/deepseek-ocr-2 | Batch | $0.0195 | n/a | $0.0195 |
| openrelay/bge-m3 | Sync | $0.013 | n/a | n/a |
| openrelay/bge-m3 | Async | $0.0104 | n/a | n/a |
| openrelay/bge-m3 | Batch | $0.0065 | n/a | n/a |

Details: https://openrelay.inc/inference-pricing

## Key pages

Links below point directly to Markdown. Drop the .md suffix for the canonical
HTML page, or request that page with Accept: text/markdown to get this same
content.

- [Rent a GPU server](https://openrelay.inc/rent-gpu.md): hourly rates, what is included, how to deploy
- [GPU catalog and pricing](https://openrelay.inc/gpus.md): every GPU, specs, live rates
- [Pricing calculator](https://openrelay.inc/gpu-pricing-calculator.md): hourly and monthly cost
- [Inference catalog](https://openrelay.inc/inference-catalog.md): OpenAI-compatible API, hosted open models
- [Become a provider](https://openrelay.inc/providers.md): earn from idle GPU capacity, paid hourly for usage
- [Bring your own capacity](https://openrelay.inc/bring-your-own-capacity.md): onboard a reserved fleet, managed orchestration, idle hours earn inference revenue
- [All markdown pages](https://openrelay.inc/sitemap.md): the full list
- [Docs](https://docs.openrelay.inc): API reference, onboarding, guides
- [Docs for LLMs](https://docs.openrelay.inc/llms.txt): full documentation index in this format
- [Blog](https://openrelay.inc/blog): announcements and engineering posts

## Comparisons

- [RunPod alternative](https://openrelay.inc/compare/runpod.md): pricing, reliability, and what each includes
- [Vast.ai alternative](https://openrelay.inc/compare/vast-ai.md): marketplace risk, failover, pricing
- [Lambda Labs pricing compared](https://openrelay.inc/compare/lambda.md): rates and availability side by side
- [AWS GPU alternative](https://openrelay.inc/compare/aws.md): EC2 and SageMaker costs compared