Skip to main content

Inference and models FAQ

Yes. Hyperbolic Inference serves text and vision models through an OpenAI-compatible chat completions API at https://api.hyperbolic.ai/inference/v1. See the Inference overview and the Quick start.
Not without changes. Requests to https://api.hyperbolic.xyz/v1 return 404. Change the base URL to https://api.hyperbolic.ai/inference/v1, use an sk_live_ API key, and copy current model IDs from the Models page. See Moving from the previous inference API.
Keys issued in the older format return 401 with the code legacy_key_unsupported. Create a new key under Settings > API Keys and use the sk_live_ key it gives you. For other error codes, see Errors.
Yes, after you verify your phone number. Verifying also adds a $1 credit. A number already verified on another Hyperbolic account doesn’t count. You can also upgrade to Pro under Settings > Plan. Requests made with an Organization API key don’t require either. See Who can use inference.
Images, yes. glm-5.3-flash, step-5-preview, minimax-m3, kimi-k3, and mimo-v2.6-flash accept images in chat completion requests and in the Model Playground. Send images as base64 data URLs. All of them except glm-5.3-flash also accept public HTTPS URLs. See Image input.Video input is experimental. step-5-preview, minimax-m3, kimi-k3, and mimo-v2.6-flash accept video through the API only, and glm-5.3-flash doesn’t accept video. The Model Playground can’t attach videos yet. See Image and video input.
No. To run those models, rent an on-demand GPU and use your own serving stack.
The Models page lists every model with its price per million tokens. See Models and pricing.To request a model that isn’t listed, email support@hyperbolic.ai with the model, roughly how many requests you send, and what you use it for.