A shorter route to capable AI

The best-value AI models. One clean API.

Ship with Chinese models through endpoints compatible with OpenAI and Anthropic.

  • No contract
  • Usage-based billing
  • OpenAI compatible
LIVE ROUTE
from openai import OpenAI

client = OpenAI(
  base_url="https://api.hubrium.com/v1",
  api_key="$HUBRIUM_KEY"
)

response = client.chat.completions.create(
  model="deepseek-v3.1",
  messages=[...] 
)

Launch catalogue

Serious models. Sensible spend.

A focused line-up for agents, code, reasoning and high-volume production work.

  • Code agents
  • Repositories
  • Tool calls
Input$0.40/ 1M tokens
Output$1.60/ 1M tokens
  • Agentic tasks
  • Coding
  • Long context
Input$0.60/ 1M tokens
Output$2.50/ 1M tokens
  • Reasoning
  • Coding
  • Agents
Input$0.60/ 1M tokens
Output$2.20/ 1M tokens

Built for production

Less gateway overhead. More room to build.

Hubrium adds the operational layer teams need without getting between you and the model.

01

One key, Chinese model families

Use one key across supported models from either compatible API format.

02

Spend you can explain

Input, cached input and output are itemised. Usage logs make every charge traceable.

03

Agent-ready by default

Streaming, JSON output and tool calling are exposed where the underlying model supports them.

Cost calculator

Price a workload before you route it.

Choose a model and adjust monthly token volume for a quick estimate.

10M

about 10K medium chat turns

5M
Estimated monthly usage$9.00

Estimate excludes taxes and optional add-ons. Billed from metered usage.

Add credit

Drop-in migration

Change the base URL. Keep your momentum.

  1. 1

    Create a key

    Open an account and generate a scoped API token.

  2. 2

    Point your SDK

    Use our base URL with your existing OpenAI or Anthropic client.

  3. 3

    Pick a model

    Route each workload to the right price-performance tier.

Your next inference call

Make it cost less—not mean less.

Start with usage-based billing. Scale when the model earns its place in your stack.

Create an accountRead the quickstart

Add prepaid credit in the console and spend it across every model. No plan, no monthly minimum, and unused credit stays refundable.