The cheapest OpenAI model, full stop. Same 272K window as GPT-5.4 at 2% of the price.
GPT-5.4 nano is OpenAI's answer to "what's the least we can charge and still be useful": $0.05 input and $0.40 output per 1M tokens — the cheapest input rate on our entire API table. ChatGPT Plus offers it unlimited, and on the API it makes per-call cost effectively a rounding error for most products.
The catch is exactly what you'd expect. Nano is for tasks you could almost write a regex for — but not quite. Classification, formatting, simple extraction, autocomplete. The moment a task needs judgment, you want mini ($0.25/$2) or full GPT-5.4 ($2.50/$15).
Within its lane, nano is genuinely good: it keeps the family's 272K context window, follows output-format instructions reliably, and is the fastest model OpenAI ships. That combination — huge window, tiny price, high speed — makes it a strong "reader" model: skim a long document and pull out the three fields you need.
The honest weakness: shallow reasoning and a tendency to take ambiguous instructions literally. Don't hand it multi-step jobs, and don't expect it to notice that your prompt contradicts itself.
| Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| GPT-5.4 nano | $0.05 | $0.40 | 272K |
| GPT-5.4 mini | $0.25 | $2 | 272K |
| GPT-5.4 | $2.50 | $15 | 272K |
| Gemini 2.5 Flash-Lite | $0.10 | $0.40 | 1M |
| DeepSeek V4-Flash | $0.14 | $0.28 | 1M |
At this price tier the rivals are Gemini 2.5 Flash-Lite ($0.10/$0.40) and DeepSeek V4-Flash ($0.14/$0.28) — both cheaper on input with 1M windows, both outside the OpenAI ecosystem. Nano wins when you're already on the GPT stack and want one-parameter downgrades for easy traffic.