AI Gateway Docs

Models

Supported models and per-token pricing on AI Gateway

The tables below list every model currently routed through AI Gateway. Prices are denominated in CNY (¥) per 1M tokens, except image models which are billed per call. Live pricing is published at the Model Marketplace.

Claude (pay-as-you-go)

Model IDInput / 1MOutput / 1MCache read / 1MCache write / 1M
claude-opus-4-7¥2.0000¥10.0000¥0.2000¥2.5000
claude-opus-4-6¥2.0000¥10.0000¥0.2000¥2.5000
claude-opus-4-5-20251101¥2.0000¥10.0000¥0.2000¥2.5000
claude-sonnet-4-6¥1.2000¥6.0000¥0.1200¥1.5000
claude-sonnet-4-5-20250929¥1.2000¥6.0000¥0.1200¥1.5000
claude-haiku-4-5-20251001¥0.4000¥2.0000¥0.0400¥0.5000

OpenAI (pay-as-you-go)

Model IDInput / 1MOutput / 1MCache read / 1MCache write / 1M
gpt-5.5¥0.2500¥1.5000¥0.0250¥0.0250
gpt-5.5-openai-compact¥0.2500¥1.5000¥0.0250¥0.0250
gpt-5.4¥0.1250¥0.7500¥0.0130
gpt-5.4-openai-compact¥0.1250¥0.7500¥0.0130
gpt-5.4-mini¥0.0380¥0.2250¥0.0040
gpt-5.3-codex-openai-compact¥0.3750¥3.0000¥0.0380
gpt-5.3-codex¥0.0880¥0.7000¥0.0090¥0.0090
gpt-5.2¥0.0880¥0.7000¥0.0090¥0.0090

Spark · iFlytek (pay-as-you-go)

Model IDInput / 1MOutput / 1MCache read / 1MCache write / 1M
gpt-5.3-codex-spark¥1.5750¥12.6000¥0.1580¥0.1580

Image models (per-call)

Model IDPrice / call
gpt-image-2¥0.100
gpt-image-2-pro¥0.100

Billing notes

  • Per-token billing — input and output tokens are priced separately. Prompt-cache hits settle at the cache read / cache write rate, materially reducing the cost of repeated context.
  • Per-call billing — image models are billed by request count, independent of input resolution or parameters.
  • Live pricing wins — group tiers and promotional discounts may differ; the authoritative source is code.b886.top/pricing. This page is a snapshot taken on 2026-05-24.

On this page