Serverless Inference
Hyperbolic’s Serverless Inference is an affordable AI inference platform designed for developers who need fast, serverless deployment of open-source models. It eliminates the complexity of managing GPU infrastructure, offering one-click deployment, low-latency inference, and full privacy control with zero data retention.
Overview
Hyperbolic supports 25+ open-source text, image, vision-language, and audio models, delivering inference at a fraction of the cost of major cloud providers while remaining fully API-compatible with OpenAI and other ecosystems.
Features
-
Zero Data Retention
All requests are stateless—your data is never stored or reused. -
Base vs. Instruct Models
- Base models are versatile completion engines, ideal for open-ended tasks.
- Instruct models are fine-tuned for direct commands and structured outputs.
Developer Tools & API Support
- Multi-Language SDKs: Generate API requests using Python, TypeScript, and cURL.
- API Playground: Test models before paying, with live adjustments for temperature, max tokens, and top-p.
- REST API: Access models via a Chat Completion-compatible REST API, with streaming support for token-by-token responses in chat applications
- Python & TypeScript: Fully OpenAI compatible, just swap your
api_keyandbase_urlto switch to Hyperbolic's models. - Gradio & HF Spaces: Deploy and interact with models using Gradio, with one-click deployment to Hugging Face Spaces for easy prototyping and shareable web interfaces.
Tiered Pricing
| Tier | RPM Limit | IP Limit | Min. Deposit | Support | Dedicated Instances | Fine-Tuning |
|---|---|---|---|---|---|---|
| Basic | 60 | 100 | $0 | No | No | No |
| Pro | 600 | 100 | ≥ $5 | No | No | No |
| Enterprise | Unlimited | Unlimited | Contact Sales | Yes | Yes | Yes |
Note: Each source IP is capped at 600 RPM to prevent DDoS.
For full details, visit app.hyperbolic.xyz/models or email [email protected].
Updated 7 days ago