Documentation
Documentation
Models
Last updated:
Choosing a model
If you're not sure where to start, use gpt-5.5, the flagship model for complex reasoning and coding. If you're optimizing for latency and cost, choose a smaller variant like gpt-5.4-mini or gpt-5.4-nano. All models use the official OpenAI request format and are available via the Responses API and the official client SDKs.
Frontier models
Prices are per 1M tokens, already 20% below official OpenAI rates.
GPT-5.5
Newflagship, best all-round balance
- Model ID
gpt-5.5- Input price
- $4.00 / MTok
- Output price
- $24.00 / MTok
- Latency
- Fast
GPT-5.5-pro
most powerful for hard tasks
- Model ID
gpt-5.5-pro- Input price
- $24.00 / MTok
- Output price
- $144.00 / MTok
- Latency
- Thorough
GPT-5.4
best price/quality
- Model ID
gpt-5.4- Input price
- $2.00 / MTok
- Output price
- $12.00 / MTok
- Latency
- Fast
GPT-5.4-mini
fast and cheap
- Model ID
gpt-5.4-mini- Input price
- $0.60 / MTok
- Output price
- $3.60 / MTok
- Latency
- Faster
GPT-5.4-nano
fastest and cheapest
- Model ID
gpt-5.4-nano- Input price
- $0.16 / MTok
- Output price
- $1.00 / MTok
- Latency
- Fastest
GPT-5.3-codex
tuned for code
- Model ID
gpt-5.3-codex- Input price
- $1.40 / MTok
- Output price
- $11.20 / MTok
- Latency
- Fast
Which model should I use?
Pick by the job, not the benchmark. Here's the OpenAI lineup mapped to common tasks — to switch, just pass the model ID in the `model` field.
| Task | Model | Why |
|---|---|---|
| Complex reasoning & coding | gpt-5.5 | Flagship, best all-round for hard problems |
| Hardest tasks, maximum capability | gpt-5.5-pro | Most powerful for demanding work |
| Everyday coding & vibe coding | gpt-5.4 | Best balance of price and quality |
| Code-tuned tasks & agents | gpt-5.3-codex | Optimized for code; pairs with Codex CLI, Cursor, Copilot |
| Fast, high-volume completions | gpt-5.4-mini | Fast and cheap |
| Cheapest, autocomplete | gpt-5.4-nano | Fastest and cheapest |
Model FAQ
Which model is best for coding?
For the hardest coding, refactoring and agentic tasks, gpt-5.5 (or gpt-5.5-pro) is strongest. gpt-5.3-codex is tuned specifically for code and works well in Codex CLI, Cursor and GitHub Copilot. For fast, cheap day-to-day work, gpt-5.4-mini and gpt-5.4-nano give the best speed-to-cost balance.
What is vibe coding, and which model should I use?
Vibe coding means building software by describing what you want in natural language and letting an AI model write and iterate on the code while you review. Use gpt-5.5 for planning and complex changes, and gpt-5.4 or gpt-5.3-codex for day-to-day implementation — connected through Cursor, Codex CLI or GitHub Copilot by pointing them at our endpoint.
Which OpenAI model is the cheapest?
gpt-5.4-nano is the cheapest and fastest, followed by gpt-5.4-mini. All prices are 20% below the official OpenAI rates — see the pricing page for the full per-model table.
How do I switch between models?
Change the model ID in the standard `model` field of your request. Everything else — your key, base URL, SDK and code — stays the same, so switching models is a one-line change.