Guides & pitfalls
Practical material on AI API relay services: how rates are calculated, how to verify a model has not been downgraded, what enterprises should check before buying, and the pitfalls we ran into while measuring these services ourselves.
Browse by topic
All articles
After you have chosen: how to keep monitoring your API provider
Choosing a site is not the end of the job. This article gives a lightweight way to keep monitoring a provider: what to p…
Configuring multiple API providers for failover
A single point of dependency is the last thing a production system should have. This article gives a failover setup that…
OpenAI protocol compatibility: chat/completions vs responses
The two endpoints look like they do the same thing, but using the wrong one gives you an immediate 404. This article exp…
Reading a relay service's response latency: one measurement is not the experience
Where the millisecond figure on the page comes from, what it can tell you, and how it is misread. Once you know the meth…
How uptime is measured: our methodology and its limits
Where our published uptime figures come from, how many samples sit behind them, what the methodology is, and what its li…
What the /v1/models endpoint is, and why it can verify a relay service
This is the easiest endpoint in the OpenAI protocol to verify, and the easiest to overlook. This article explains what i…
Sharing one API key across a team: how to manage it
A shared key creates two problems: usage cannot be attributed, and incidents cannot be traced. This article sets out the…
Customer support and chat: balancing reliability against concurrency
User-facing chat tolerates no interruption. This article covers the selection criteria for that scenario, how to estimat…
Batch data processing: how to use a low-rate site correctly
Batch jobs tolerate latency and are extremely cost-sensitive, which makes them the natural home of a low-rate site. But…
Choosing for code completion: why latency beats the rate multiplier
Every extra 300 milliseconds of completion latency in your editor breaks your flow once more. This article explains why…
Choosing an API as an individual developer: work backwards from monthly usage
The biggest mistake individual developers make is starting from unit price. This article gives a method that works backw…
Mainland relay sites vs overseas platforms: payment, latency, compliance
They get compared constantly, but they suit very different situations. This article covers payment, network, pricing mod…
Supporting data & tools
The articles explain the method; the pages below give you measured data and prices to check it against. Reading them together is easier — once you know how to tell whether a rate is real, you can look the official price up in the price list.