# OpenAI

> ZopNight registers 5 OpenAI models by default through the AI Gateway: gpt-4o and gpt-4-turbo as strong tiers, gpt-4o-mini as the cheap tier, and o1 with o1-mini for reasoning. Spend is attributed per request, per team and per key, against hard USD budgets you set.

Source: https://zop.dev/integrations/ai-gateway/openai
Updated: 2026-08-20

---

## Common questions

### Which OpenAI models are registered by default?

Five: gpt-4o, gpt-4-turbo and o1 in the strong tier, gpt-4o-mini and o1-mini in the cheap tier.

### Should I route everything to o1?

No. The o1 pair are reasoning models, which makes them the expensive right answer for problems that need them and a poor default for everything else. Route them deliberately rather than making one the org default.

### What happens when a team's OpenAI key hits its budget?

The key stops working. Virtual keys carry a hard USD cap rather than an alert, each with its own model allow-list, so a leaked key is contained to one team's budget instead of the org's.

## GPT-4o, GPT-4 Turbo and the o1 reasoning pair

Five, split by tier so routing has somewhere to send each class of request:

- **Strong**: `gpt-4o`, `gpt-4-turbo`, `o1`
- **Cheap**: `gpt-4o-mini`, `o1-mini`

The `o1` pair are reasoning models, and worth routing deliberately: they are the expensive
option for problems that need them and a poor default for everything else.

## OpenAI calls do not traverse ZopNight

ZopNight never sits in your request path. The gateway runs on hosted LiteLLM; ZopNight
configures it, meters it and enforces the budget, but your prompts and completions do not
traverse our infrastructure. That distinction is why per-request cost attribution is possible
without us reading request bodies.

## Virtual OpenAI keys with hard USD caps

Each team gets a virtual key with a hard USD budget and its own model allow-list. When a key
hits its cap it stops working. This is a limit, not an alert. Keys rotate and revoke
independently, so a leaked key is contained to one team's budget rather than the org's.

## Cost-saver, Balanced and Quality-first for OpenAI

Three postures decide which model serves a request: **Cost-saver**, **Balanced** and
**Quality-first**. Requests are classified into four semantic categories (coding, analysis and
reasoning, extraction and classification, chit-chat) and routed accordingly, so cheap models
take the traffic that does not need an expensive one.

A semantic cache sits in front, with strictness settings from strict (0.98 similarity) to loose
(0.90). On repetitive workloads the cache is usually a larger saving than the routing itself.