MIXERBOX CLOUD

The Lowest-Cost AI Cloud

// One endpoint · every frontier model · pay per output token.

OpenAI-compatible /v1/chat/completions. Bring your existing SDK, point at our base URL, ship.

openai-python v1
from openai import OpenAI

client = OpenAI(
  base_url="https://cloud.mixerbox.com/v1",
  api_key="sk-mb-…",
)

resp = client.chat.completions.create(
  model="kimi-k2.5",
  messages=[{"role": "user", "content": "hi"}],
)

Swap the base URL — that's the whole integration.

Models from every provider, one API

OpenAI Google Alibaba MoonshotAI Minimax Z.AI Nvidia

Built for builders, not budgets

A drop-in OpenAI-compatible endpoint priced below AWS Bedrock and the hyperscalers. Pay per token, never per seat.

Lowest cost

Output priced below AWS Bedrock and most hyperscalers. Provider list price plus a flat markup, posted alongside ours on the /models page.

Unified API

OpenAI-compatible /v1/chat/completions. Swap your base URL, point at any model. No SDK rewrite.

Pay-as-you-go

Top up your balance once. Pay per token. No subscription, no per-seat fee, no minimum commitment.

Auto top-up

Save a card and we'll auto-charge when your balance falls below a threshold you set. Set it once, never get throttled.

Stream-first

SSE streaming on every model. Same protocol, no special casing.

Multi-modal

Beyond chat: TTS, STT, embeddings, and vision input on supported models — all through the same /v1/ endpoint.

Three lines, every model

Point your existing OpenAI client at us. Nothing else changes.

/docs → full API reference
curl https://cloud.mixerbox.com/v1/chat/completions \
  -H "Authorization: Bearer sk-mb-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.2",
    "messages": [{"role":"user","content":"Hello"}],
    "stream": true
  }'

Pay only for what you use

No subscription, no minimum, no surprise. Provider price + 5% markup, billed per token. Top up any amount, anytime.

See pricing breakdown →

Frequently asked

Is this OpenAI-compatible? +

Yes. /v1/chat/completions, /v1/embeddings, /v1/audio/transcriptions — exact same shape as the OpenAI API. Most clients work by changing only the base URL.

How is this different from OpenRouter? +

Functionally similar — unified API across LLMs. We differ in Traditional Chinese-first support, local NT$ invoicing, and being a Taiwan-resident company for B2B contracts.

What's the markup? +

Currently 5% on top of provider list price. We post raw provider prices alongside ours on the /models page so you can verify.

Can I get a refund? +

Unused credit can be refunded for 30 days from the top-up date, minus payment processing fees (~3%). After 30 days credit is non-refundable.

Do you log my prompts? +

We log metadata (model, token count, latency) for billing. Prompt/response content is not logged by default. Enterprise plans can opt into full logging for debugging.

Auto top-up? +

Coming soon. Save a card and we'll auto-charge when balance falls below your threshold. Available shortly after launch.