One API.
Every frontier model.
MAX API is a unified gateway to the world's leading AI models. We give teams one dependable way to reach them — one key, one endpoint, one bill — so they can spend their time building products instead of plumbing.
- models in the catalog
- 151
- models in the catalog
- model vendors
- 11
- model vendors
- model families
- 12
- model families
- official list prices
- 100%
- official list prices
Make frontier AI simple, dependable and fairly priced for every team
Our mission
Remove the friction between great models and great products. Every team should be able to use the best model for each job — without negotiating a dozen contracts, maintaining a dozen SDKs or reconciling a dozen invoices.
Our vision
A neutral, trusted access layer for AI: the infrastructure companies rely on to adopt new models the day they ship, move freely between them, and run them in production with confidence.
The layer that makes model complexity disappear
Every model vendor has its own API, accounts, quotas and failure modes. MAX API sits between your application and all of them, so you integrate once and get the whole market.
One API for the top models
OpenAI- and Anthropic-compatible endpoints give you GPT, Claude, Gemini, DeepSeek, Qwen and more with a single key. Switching models is a one-line change.
Smart routing
Each request is dispatched in real time to a healthy upstream channel, weighing availability, error rates and latency — with automatic failover when something goes wrong.
Official list prices
Pay per token at the vendors' published list prices. No subscriptions, no seat fees and no minimum spend — every price is public in our catalog.
Enterprise-grade stability
Multi-channel redundancy, circuit breaking and continuous health monitoring keep production traffic flowing, even when an individual provider has a bad day.
Built for production traffic, not demos
Reliability is engineered into every hop of the request path — from the moment a request reaches our gateway to the last streamed token.
Multi-provider redundancy
Popular models are served through several independent upstream channels, so no single provider, account or region becomes a single point of failure.
Automatic scheduling & circuit breaking
Channels are continuously scored on health and performance. Failing or slow channels are taken out of rotation automatically, and retries land on healthy ones.
Low latency
End-to-end streaming and a lean gateway path keep our overhead minimal, so time-to-first-token is governed by the model — not by the gateway.
Observable request path
Every request is metered and traceable — model, tokens, status and latency — so usage and spend are transparent in your console.
Your data stays yours
We process what is needed to route and fulfil your request — and nothing more. Our practices are set out in full in our Privacy Policy.
Read the full details in our Privacy Policy and Terms of Service.
No training on your data
We do not use your prompts, files or outputs to train our own models or any third-party model.
Minimal retention
API content is processed transiently to fulfil each request and is not persistently stored by default. Operational logs focus on metadata such as model, token counts, status and latency.
Encrypted in transit
Traffic between your application, our gateway and model vendors is encrypted with TLS.
Access control
Internal access is role-based and least-privilege. API keys are stored with cryptographic protection and shown in full only once, when they are created.
Dedicated groups & pricing
Dedicated routing groups for your workloads, with volume-based discounts agreed for your usage.
Negotiated SLAs
Availability and support commitments agreed to match the requirements of your production systems.
Hands-on technical support
An account manager and engineers who help with integration, model selection, capacity planning and incident response.
Clear usage & spend
One balance across every vendor, with usage broken down by key and model in your console.
A partner for teams running AI at scale
Beyond self-serve access, we work directly with companies that depend on AI in production.
Models from the world's leading labs
151 models from 11 vendors, available through one API.
Logos and trademarks belong to their respective owners. They are shown to indicate model availability on MAX API and do not imply endorsement or partnership.
Principles we don't trade away
Reliability
Production traffic deserves production engineering. We measure ourselves by your uptime.
Transparency
Public prices, clear policies and usage you can audit request by request.
Neutrality
We don't push a favourite model. We help you use the right one for each job.
Simplicity
One key, one endpoint, one bill. Complexity is our job, not yours.
From one gateway to a model platform
- Stage 01
The problem
It started with our own frustration: every new model meant a new account, a new SDK and a new invoice.
- Stage 02
One gateway
We built a unified, OpenAI-compatible gateway with multi-channel routing and automatic failover.
- Stage 03Now
A growing catalog
Anthropic-compatible endpoints, image and embedding models, and a public catalog at official list prices — today 151 models from 11 vendors.
- Stage 04Next
What's next
Deeper enterprise capabilities, richer observability and new modalities as the frontier moves.
Let's talk
Whether you are evaluating MAX API or scaling an existing deployment, we are here to help.
Existing customers
Reach out through your account manager for pricing, capacity, SLAs or technical support.
Planning an enterprise rollout?
Get in touch through your MAX API account manager — we will set up a dedicated group, pricing and support for your team.
- Hong Kong
Start building with MAX API
Browse the catalog, compare prices and make your first request in minutes.