One route.
More intelligence.

Zosma Router gives your products a clean, reliable path to production AI models—with one key, predictable daily limits, and no infrastructure to babysit.

One API key

OpenAI-compatible

Clear capacity

Daily credits

Built for shipping

Simple integration

Routing manifest
LIVE / ZR-01
POST /v1/chat
MiMo V2.5
DeepSeek V4 Flash 0731
GPT 5.6 Luna
GPT 5.6 Terra
One key. Every route.99.9% ready
Weighted routingExplore specs

Model manifest / 01

Choose the intelligence your product needs.

Every model uses the same API shape. Request size shows how many daily credits one request consumes.

DeepSeek V4 Flash 0731

deepseek/deepseek-v4-flash-0731

Request 1×
Context 1.05M
Input Text
Training No training
Retention 0 days · ZDR
A 5× request uses five daily credits. Model availability can evolve as the platform grows.

Data policy / 02

Clear about what happens to your data.

Privacy should be a product surface, not a footnote. See retention at the model level before you ship.

No training

Your prompts and outputs are not used to train models.

ZDR models

MiMo and DeepSeek routes retain request data for 0 days.

30-day GPT retention

GPT routes retain data for 30 days for service operation.

Policies are shown alongside every model

Plans / 03

Capacity that grows with you.

All plans include every listed ZosmaAI model and one consistent API surface.

Hobby

For personal projects

499/ month

500 daily credits

500 × 1× requests · 166 × 3× requests

  • All ZosmaAI models
  • OpenAI-compatible API
  • Usage visibility in console
Most popular

Pro

For production apps

999/ month

1,000 daily credits

1,000 × 1× requests · 333 × 3× requests

  • All ZosmaAI models
  • OpenAI-compatible API
  • Usage visibility in console

Business

For teams

1,999/ month

2,000 daily credits

2,000 × 1× requests · 666 × 3× requests

  • All ZosmaAI models
  • OpenAI-compatible API
  • Usage visibility in console

Start free — 30-day trial with 100 daily requests

No card required. Full access to every ZosmaAI model on the router.

Start with a plan, then scale when your traffic says it is time.