GPT Series

The OpenAI GPT series is the general-purpose mainline, covering flagship chat, lightweight batch, and code-specific tiers.

Models for Sale

Model	Positioning
`gpt-5.5`	Flagship mainline, strongest overall
`gpt-5.5-high` / `-low` / `-xhigh`	Same model with different reasoning intensity tiers
`gpt-5.6-luna` / `-sol` / `-terra`	Three stylistic branches of the 5.6 generation
`gpt-5.4-mini`	Lightweight tier, cost-effective choice for batch tasks
`gpt-5.3-codex`(`-high`)	Code-specific tier

See real-time pricing on Model Marketplace.

resp = client.chat.completions.create(
    model="gpt-5.4-mini",
    messages=[{"role": "user", "content": "把这段话改写得更正式:……"}],
)

Primary product chat: gpt-5.5; for more stable complex reasoning, switch to gpt-5.5-high, and for the toughest problems use -xhigh;
Batch pipelines (summarization/classification/paraphrasing): gpt-5.4-mini, with a cost difference of an order of magnitude at large scale;
Code generation/completion: gpt-5.3-codex, together with the Codex toolchain;
The three 5.6 branches (luna/sol/terra) have different stylistic orientations; it is recommended to compare them on the same prompt and then fix one.