State-of-the-art AI Models
Why Run Inference on Hyperbolic?
Serverless API powered by Flux
Flux is Hyperbolic’s inference platform, built on Forge, our unified compute and orchestration layer. Manage model access and token-based usage alongside GPU compute through one Hyperbolic account.
Scalable inference capacity
Access inference without reserving or managing GPUs. Requests are subject to model availability, capacity, and rate limits.
Affordable pricing
Pay only for the tokens you use, with clear model-specific rates and no hidden fees or long-term commitments.
Leading models
Run leading AI models through one OpenAI-compatible API, without provisioning GPUs or managing serving infrastructure. Build prototypes, internal tools, and production applications with models including DeepSeek V4.1 Flash, GLM 5.3 Flash, Step 5 Preview, and MiMo-V2.6-Flash.
Pricing
Our transparent, usage-based pricing model keeps things predictable. Rates are displayed upfront, billing is straightforward, and you can choose the setup that matches how you run inference today.
All models
Every model available through our inference API, billed per token.
11 models
?
MiMo-V2.6-Flash
mimo-v2.6-flash
GLM 5.3
glm-5.3
Kimi K3
kimi-k3
GLM 5.3 Flash
glm-5.3-flash
Nemotron 3 Ultra
nemotron-3-ultra-550b-a55b
DeepSeek V4 Flash 0731
deepseek-v4-flash-0731
MiniMax M3
minimax-m3
Hy4 Preview
hy4-preview
DeepSeek V4.1 Flash
deepseek-v4.1-flash
Step 5 Preview
step-5-preview
Qwen3.8 27B
qwen3.8-27b
Open-model inference, powered by Forge
Run leading AI models through one OpenAI-compatible API, without provisioning GPUs or managing serving infrastructure. Flux brings model access, streaming responses, and token-based billing into your Hyperbolic account. Built on Forge, it connects inference with the orchestration platform behind our GPU services.
Access leading models, including GLM-5.3-Flash and DeepSeek V4.1 Flash
Use the OpenAI SDK with your Hyperbolic API key and endpoint
Stream responses for AI applications, agents, and evaluation workflows
Pay by token usage at model-specific rates with prepaid credits
Manage inference and GPU compute through one Hyperbolic account
Made for Making
Dedicated Hosting
Accelerating Developer Access to Open-Source AI
“
Finding a host for the particular model we've been looking to use wasn't easy — Hyperbolic was the only platform that had it ready to go. Not only has the performance been outstanding, but their pricing absolutely crushes the major competitors. On top of that, the Hyperbolic founders provide the best customer support we’ve experienced, always going above and beyond to solve our needs. Partnering with them has been a huge win for us.
:format(webp))
Inference FAQs
Flux is Hyperbolic’s inference service for running open AI models through one OpenAI-compatible API. Build applications, stream responses, and manage token-based usage without provisioning GPUs or operating model-serving infrastructure.
The Flux model catalog includes GLM-5.3-Flash, DeepSeek V4.1 Flash, Kimi K3, MiniMax M3, MiMo-V2.6-Flash, and other models. Check the current catalog for available models, pricing, and supported parameters.
Yes. Flux supports OpenAI-compatible Chat Completions and streaming responses. Configure the OpenAI SDK with your Hyperbolic API key, set the base URL to https://api.hyperbolic.ai/inference/v1, and choose a supported model ID. Model-specific parameters and reasoning behavior may differ.
Flux uses token-based pricing, with rates set for each model. Input, cached-input, and output tokens may have different rates where supported. Reasoning tokens are billed as output tokens. Usage is deducted from your prepaid Hyperbolic balance.
Use a Pro account or an Organization with prepaid credits, create an API key in the Hyperbolic console, and connect through the OpenAI SDK or REST API. Organization API keys use the organization’s balance. Requests are declined when the balance runs out.
Yes. Flux streams Chat Completions through server-sent events, allowing applications to display responses incrementally. Reasoning models may generate reasoning before final-answer text appears, and reasoning response formats vary by model.
No. Hyperbolic manages the model-serving infrastructure behind the API. You select a supported model and send requests, without configuring GPU instances, installing serving software, or maintaining model deployments.
An inference API provides access to supported models with token-based billing. Dedicated GPU compute gives you control over the deployment environment, model weights, and serving configuration. The right choice depends on your customization needs, traffic, and GPU utilization.
Flux is built on Forge, Hyperbolic’s infrastructure orchestration platform. Forge integrates compute across providers and clusters, supports virtualization, and dynamically allocates GPU capacity across workloads.
Flux requires an API key that starts with sk_live_. If your key uses the older format, create a new one in the Hyperbolic console. Applications using the earlier inference API also need the new base URL and a supported model ID.